MSU Quality Measurement Tool: Metrics information

MSU Graphics & Media Lab (Video Group)


Metrics Info


This page describles metrics effectively implemented in MSU VQMT

Download free | Purchase | More info...

Peak signal-to-noise ratio (PSNR)

General info

Metric type:
full-reference image metric
Value range:
(completely different) 0..100 (similar to original)
Value interpretation:
bigger is better quality
MSU VQMT implementaions:
cpu multithreaded
MSU VQMT visualization:
pixel-wise
Available colorspaces:
R, G, B, Y, U, V, L, RGB, YUV
Output values:
metric value
Aggregated values:
standard set , total PSNR
MSU VQMT usages:
-metr psnr [over <color components>]
External references:
Wikipedia

Algorithm description

Metric depends only on difference of original and distorted, and more preciesly, only on -norm of this difference (see MSE). Unlike MSE, metric has logrithmic scale and can be calculated using the following formula:

where MaxErr - maximum possible absolute value of color component (MaxErr=1 in VQMT), w - video width, h - video height.

Total PSNR is aggregated value, that considers all processed frames as a single huge image and then calculates PSNR. Total PSNR takes into account sum suqared distortion on all frames, and doesn't distinguish situation where all distorion on one frame from situation where it is distributed across all frames. While arithmetic mean aggregated value depends from geometric mean of MSE's of each frames, Total PSNR depends on arithmetic mean of them.

In MSU VQMT you can calculate PSNR for all YUV and RGB components and for L component of LUV color space. Also, since VQMT 12 you can calculate over all YUV space or all RGB space, achieving a single value for 3 components. PSNR metric is easy and fast to calculate, but sometimes it is not appropriate to human's perception.

Benchmark

Implementation System & settings Resulution FPS Sec. per frame
VQMT 13 default-Y 8 cores HD 720p 492.737 0.002
VQMT 13 default-Y 1 core enabled HD 720p 477.477 0.002
VQMT 13 default-Y 8 cores FullHD 1080p 213.148 0.005
VQMT 13 default-Y 1 core enabled FullHD 1080p 213.664 0.005
VQMT 13 default-Y 8 cores UHD 4k 2160p 48.946 0.020
VQMT 13 default-Y 1 core enabled UHD 4k 2160p 50.437 0.020
VQMT 13 default-YUV 8 cores HD 720p 365.428 0.003
VQMT 13 default-YUV 1 core enabled HD 720p 276.738 0.004
VQMT 13 default-YUV 8 cores FullHD 1080p 161.647 0.006
VQMT 13 default-YUV 1 core enabled FullHD 1080p 134.382 0.007
VQMT 13 default-YUV 8 cores UHD 4k 2160p 37.379 0.027
VQMT 13 default-YUV 1 core enabled UHD 4k 2160p 33.964 0.029
measurements were done on VQMT 12 PRO for Windows. Can vary depending on system configuration, input format and other factors

Example

Choose example:
Original
LQ H264, PSNR-Y=32.91
MSU VQMT visualization
Original
Blurring, PSNR-Y=32.31
MSU VQMT visualization
Original
Random points, PSNR-Y=37.97
MSU VQMT visualization
Original
Luminance shift, PSNR-Y=42.11
MSU VQMT visualization
Original
JPEG Q=2, PSNR-Y=26.62
MSU VQMT visualization
Original
JPEG Q=5, PSNR-Y=29.46
MSU VQMT visualization
Original
JPEG Q=10, PSNR-Y=33.01
MSU VQMT visualization
Original
JPEG Q=15, PSNR-Y=34.76
MSU VQMT visualization
Original
JPEG Q=20, PSNR-Y=36.01
MSU VQMT visualization
Original
JPEG Q=40, PSNR-Y=38.83
MSU VQMT visualization
Original
JPEG Q=80, PSNR-Y=43.31
MSU VQMT visualization
Original
LQ H264, PSNR-RGB=32.47
MSU VQMT visualization
Original
Blurring, PSNR-RGB=32.23
MSU VQMT visualization
Original
Random points, PSNR-RGB=37.93
MSU VQMT visualization
Original
Luminance shift, PSNR-RGB=42.11
MSU VQMT visualization
Original
JPEG Q=2, PSNR-RGB=25.54
MSU VQMT visualization
Original
JPEG Q=5, PSNR-RGB=27.18
MSU VQMT visualization
Original
JPEG Q=10, PSNR-RGB=31.12
MSU VQMT visualization
Original
JPEG Q=15, PSNR-RGB=32.75
MSU VQMT visualization
Original
JPEG Q=20, PSNR-RGB=34.84
MSU VQMT visualization
Original
JPEG Q=40, PSNR-RGB=38.00
MSU VQMT visualization
Original
JPEG Q=80, PSNR-RGB=42.24
MSU VQMT visualization
Choose example:
Original
LQ H264, PSNR-Y=22.31
MSU VQMT visualization
Original
Blurring, PSNR-Y=26.61
MSU VQMT visualization
Original
Random points, PSNR-Y=36.74
MSU VQMT visualization
Original
Luminance shift, PSNR-Y=42.28
MSU VQMT visualization
Original
JPEG Q=2, PSNR-Y=24.59
MSU VQMT visualization
Original
JPEG Q=5, PSNR-Y=27.05
MSU VQMT visualization
Original
JPEG Q=10, PSNR-Y=29.77
MSU VQMT visualization
Original
JPEG Q=15, PSNR-Y=31.32
MSU VQMT visualization
Original
JPEG Q=20, PSNR-Y=32.33
MSU VQMT visualization
Original
JPEG Q=40, PSNR-Y=34.80
MSU VQMT visualization
Original
JPEG Q=80, PSNR-Y=39.22
MSU VQMT visualization
Original
LQ H264, PSNR-RGB=22.14
MSU VQMT visualization
Original
Blurring, PSNR-RGB=26.58
MSU VQMT visualization
Original
Random points, PSNR-RGB=36.73
MSU VQMT visualization
Original
Luminance shift, PSNR-RGB=42.29
MSU VQMT visualization
Original
JPEG Q=2, PSNR-RGB=23.97
MSU VQMT visualization
Original
JPEG Q=5, PSNR-RGB=26.10
MSU VQMT visualization
Original
JPEG Q=10, PSNR-RGB=28.66
MSU VQMT visualization
Original
JPEG Q=15, PSNR-RGB=30.53
MSU VQMT visualization
Original
JPEG Q=20, PSNR-RGB=31.68
MSU VQMT visualization
Original
JPEG Q=40, PSNR-RGB=34.19
MSU VQMT visualization
Original
JPEG Q=80, PSNR-RGB=38.46
MSU VQMT visualization
Choose example:
Original
LQ H264, PSNR-Y=31.34
MSU VQMT visualization
Original
Blurring, PSNR-Y=24.24
MSU VQMT visualization
Original
Random points, PSNR-Y=26.34
MSU VQMT visualization
Original
Luminance shift, PSNR-Y=42.11
MSU VQMT visualization
Original
JPEG Q=2, PSNR-Y=23.97
MSU VQMT visualization
Original
JPEG Q=5, PSNR-Y=24.92
MSU VQMT visualization
Original
JPEG Q=10, PSNR-Y=26.15
MSU VQMT visualization
Original
JPEG Q=15, PSNR-Y=26.36
MSU VQMT visualization
Original
JPEG Q=20, PSNR-Y=26.46
MSU VQMT visualization
Original
JPEG Q=40, PSNR-Y=26.62
MSU VQMT visualization
Original
JPEG Q=80, PSNR-Y=26.71
MSU VQMT visualization
Original
LQ H264, PSNR-RGB=30.39
MSU VQMT visualization
Original
Blurring, PSNR-RGB=14.57
MSU VQMT visualization
Original
Random points, PSNR-RGB=14.61
MSU VQMT visualization
Original
Luminance shift, PSNR-RGB=42.10
MSU VQMT visualization
Original
JPEG Q=2, PSNR-RGB=15.10
MSU VQMT visualization
Original
JPEG Q=5, PSNR-RGB=14.51
MSU VQMT visualization
Original
JPEG Q=10, PSNR-RGB=14.69
MSU VQMT visualization
Original
JPEG Q=15, PSNR-RGB=14.66
MSU VQMT visualization
Original
JPEG Q=20, PSNR-RGB=14.68
MSU VQMT visualization
Original
JPEG Q=40, PSNR-RGB=14.67
MSU VQMT visualization
Original
JPEG Q=80, PSNR-RGB=14.65
MSU VQMT visualization

Legacy notes

Video Multimethod Assessment Fusion (Netflix VMAF)

General info

Metric type:
full-reference temporal metric
Value range:
(completely different) 0..100 (similar to original)
-∞..∞ if truncation to 0..100 is off
dependent on model in case of custom model
Value interpretation:
bigger better quality, value 100 does not mean that the images match pixel by pixel
MSU VQMT implementations:
CPU multithreaded
OpenCL (since VQMT 13)
MSU VQMT visualization:
block-wise (for VMAF visualization),
pixel-wise (for ADM, VIF, ANSNR visualisation)
Available colorspaces:
Y
Output values:
metric value,
bagging values (if on),
confidence intervals (if on),
values of elementary features (if on)
Aggregated values:
standard set
MSU VQMT usages:
-metr vmaf [-dev <OpenCL device>]
External links:
Original paper

Algorithm description

VMAF is modern reference metric developed by Netflix in cooperation with the University of Southern California. VQMT has full support of VMAF with multiple configuration switches. MSU VQMT support the following VMAF models: VMAF 0.60, VMAF 0.61 (2k, 4k), VMAF 0.62 (2k, 4k), VMAF 0.63 (2k), also, you can compute phone model and elementary features of VMAF. You can use custom model in pkl format with VQMT.

VMAF consist of 4 features (ADM, VIF, Motion, ANSNR) and 35 elementary features, but VMAF models uses only 6 of them: adm2, motion2, vif_scale0, vif_scale1, vif_scale2, vif_scale3. VMAF applies an SVM model to this set of features, which depends on current settings. After applying SVM, the value is clipped to interval 0..100 by default. Motion feature is the only temporal feature, it consider adjacent frames. To calculate VMAF value for current frame it is needed to use the previous frame and the next frame.

VMAF can also compute confidence intervals by applying multiple models and calculating standard deviation of result. VMAF has models, that aimed for 4k and 2k. By default, VQMT will automatically select the correct model by the resolution of input video.

Since VQMT 13 VMAF has a real block-wise visualization, which computes individual VMAF value for each 16x16 block of image. You also can visualize every feature besides motion (ADM, VIF, ANSNR).

MSU VQMT parameters description

Model preset
  • Description. Choose built-in model or 'custom' for loading model from file. Built-in models:
    • default - VMAF default behaviour:
      1. VMAF v0.6.1 for running without confidence interval and per-model values
      2. VMAF v0.6.1 4k for previous case if applying 4k model
      3. VMAF v0.6.3 for running with confidence interval or per-model values
      4. VMAF v0.6.2 4k for previous case if applying 4k model (NOTE: no v0.6.3 for 4k)
    • vmaf_v061 - Netflix model VMAF v0.6.1 (2k or 4k)
    • vmaf_v062 - Netflix model VMAF v0.6.2 (2k or 4k), supports confidence interval
    • vmaf_v063 - Netflix model VMAF v0.6.3 (only 2k), supports confidence interval
    • all_models - vmaf_v061 - vmaf_v063 computed sumultaneously
    • basic_features - view only basic features from VMAF. Model will not be applied
    • standard_features - features that is used in VMAF v0.6.1 and VMAF score (2k or 4k)
    • all_features - view all features from VMAF. Model will not be applied
    • all - all feature and next models:
      1. VMAF v0.6.1 (2k or 4k)
      2. VMAF v0.6.2 (2k or 4k)
      3. VMAF v0.6.3
  • Default value. default
  • Usage. -set "model_preset=<value>", where <value> can be:
    default, vmaf_v061, vmaf_v062, vmaf_v063, vmaf_v060, all_models, basic_features, standard_features, all_features, all, custom.
Custom model (*.pkl)
  • Description. You can specify path to *.pkl file here (or multiple ;-separated *.pkl files). Model file should be placed near pkl file. NOTE: this only means if preset is set to 'custom'
  • Default value. empty
  • Usage. -set "custom_model_files=<value>", where <value> can be:
    any string.
4k
  • Description. Selection 4k model policy:
    • auto - select 4k if exists suitable model and input video is 4k
    • forced_2k - always 2k model
    • forced_4k - 4k if exsists: VMAF v0.6.1-2

    NOTE: this param does not affects custom model

  • Default value. auto
  • Usage. -set "4k=<value>", where <value> can be:
    auto, forced_2k, forced_4k
Confidence interval
  • Description. Turn on additional VMAF features: 95%-confidence interval output and other statistical information
  • Default value. false
  • Usage. -set "confidence_interval=<value>", where <value> can be:
    true, false
Per-model values
  • Description. Output values for all bootstrap models if confidence interval is on
  • Default value. false
  • Usage. -set "permodel_values=<value>", where <value> can be:
    true, false
Visualize algorithm (if on)
  • Description. If visualization turned on you can select feature to visualize. It's impossible to calculate distribution of real VMAF value, so you can only visualize one of supposed features
  • Default value. default
  • Usage. -set "visualize_alg=<value>", where <value> can be: default, adm, ansnr, vif
Use phone model
  • Description. Turn on postprocessing of metric value that produces more precise results for handheld devices. Select 'both' to see both results with and without postprocessing
  • Default value. no
  • Usage. -set "phone_model=<value>", where <value> can be:
    no, yes, both
Disable clipping values
  • Description. Turn off clipping value to range set by model (0..100 for example)
  • Default value. false
  • Usage. -set "disable_clip=<value>", where <value> can be:
    true, false

Benchmark

Implementation System & settings Resulution FPS Sec. per frame
VQMT 13 default-Y 8 cores HD 720p 492.737 0.002
VQMT 13 default-Y 1 core enabled HD 720p 477.477 0.002
VQMT 13 default-Y 8 cores FullHD 1080p 213.148 0.005
VQMT 13 default-Y 1 core enabled FullHD 1080p 213.664 0.005
VQMT 13 default-Y 8 cores UHD 4k 2160p 48.946 0.020
VQMT 13 default-Y 1 core enabled UHD 4k 2160p 50.437 0.020
VQMT 13 default-YUV 8 cores HD 720p 365.428 0.003
VQMT 13 default-YUV 1 core enabled HD 720p 276.738 0.004
VQMT 13 default-YUV 8 cores FullHD 1080p 161.647 0.006
VQMT 13 default-YUV 1 core enabled FullHD 1080p 134.382 0.007
VQMT 13 default-YUV 8 cores UHD 4k 2160p 37.379 0.027
VQMT 13 default-YUV 1 core enabled UHD 4k 2160p 33.964 0.029
measurements were done on VQMT 12 PRO for Windows. Can vary depending on system configuration, input format and other factors

Example

Choose example:
Original
LQ H264, VMAF-Y=40.35
MSU VQMT visualization
Original
Blurring, VMAF-Y=10.85
MSU VQMT visualization
Original
Random points, VMAF-Y=96.37
MSU VQMT visualization
Original
Luminance shift, VMAF-Y=98.13
MSU VQMT visualization
Original
JPEG Q=2, VMAF-Y=21.25
MSU VQMT visualization
Original
JPEG Q=5, VMAF-Y=34.29
MSU VQMT visualization
Original
JPEG Q=10, VMAF-Y=57.95
MSU VQMT visualization
Original
JPEG Q=15, VMAF-Y=69.38
MSU VQMT visualization
Original
JPEG Q=20, VMAF-Y=77.03
MSU VQMT visualization
Original
JPEG Q=40, VMAF-Y=87.56
MSU VQMT visualization
Original
JPEG Q=80, VMAF-Y=94.37
MSU VQMT visualization
Choose example:
Original
LQ H264, VMAF-Y=3.69
MSU VQMT visualization
Original
Blurring, VMAF-Y=4.09
MSU VQMT visualization
Original
Random points, VMAF-Y=95.91
MSU VQMT visualization
Original
Luminance shift, VMAF-Y=97.01
MSU VQMT visualization
Original
JPEG Q=2, VMAF-Y=30.92
MSU VQMT visualization
Original
JPEG Q=5, VMAF-Y=43.10
MSU VQMT visualization
Original
JPEG Q=10, VMAF-Y=62.78
MSU VQMT visualization
Original
JPEG Q=15, VMAF-Y=73.76
MSU VQMT visualization
Original
JPEG Q=20, VMAF-Y=78.95
MSU VQMT visualization
Original
JPEG Q=40, VMAF-Y=88.32
MSU VQMT visualization
Original
JPEG Q=80, VMAF-Y=94.38
MSU VQMT visualization
Choose example:
Original
LQ H264, VMAF-Y=70.33
MSU VQMT visualization
Original
Blurring, VMAF-Y=5.08
MSU VQMT visualization
Original
Random points, VMAF-Y=87.23
MSU VQMT visualization
Original
Luminance shift, VMAF-Y=100.00
MSU VQMT visualization
Original
JPEG Q=2, VMAF-Y=47.35
MSU VQMT visualization
Original
JPEG Q=5, VMAF-Y=58.73
MSU VQMT visualization
Original
JPEG Q=10, VMAF-Y=72.36
MSU VQMT visualization
Original
JPEG Q=15, VMAF-Y=77.01
MSU VQMT visualization
Original
JPEG Q=20, VMAF-Y=79.93
MSU VQMT visualization
Original
JPEG Q=40, VMAF-Y=83.90
MSU VQMT visualization
Original
JPEG Q=80, VMAF-Y=86.39
MSU VQMT visualization

Naturalness Image Quality Evaluator (NIQE)

General info

Metric type
no-reference image metric
Value range
(very natural) 0..∞ (not natural)
Value interpretation
bigger metric values - better quality (naturalness)
normal metric values are in range about 3..20. Also, there can be values NAN and 0, which are abnormal and should be considered as symptom of not natural image
MSU VQMT implementations
CPU multithreaded
OpenCL (since VQMT 13)
MSU VQMT visualization
block-wise
Available colorspaces
Y
Output values
metric value
Aggregated values
standard set , NIQE mean
MSU VQMT usages
-metr niqe [-dev <OpenCL device>]
External links
original paper (A. Mittal, R. Soundararajan and A. C. Bovik),
MSU paper

Algorithm description

NIQE performs feature extraction for every 96x96 block of image on 2 scales. Than it computes correlations of features between all blocks and applies leaned model. For more details, please refer to original paper. Since version 13 VQMT can build a block-wise visualization showing contribution of each block to the final result.

This metric has a special aggregated value NIQE mean, it takes a weighted mean filtering abnormal metric values and taking suspicies values with low weight. You can turn this mode using metric settings. Also, please see our paper.

This metric only applicable to filmed scenes. Metric can produce inadequate result if some graphics is on it (including credits, subtitles, etc.). Please, run NIQE excluding all rendered scenes, they can fatally spoil average value. Also, this metric can produce bad result on scenes containing noisy objects, like sand or grass, however on scenes with big constant areas, like monotonic sky. In common case normal metric results lies in the interval 3..20.

Sometimes, metric shows better result for the compressed image and this correlates with human perception. Compressed image not always is perceived as worse. It can occur for example in case of noisy images (if the noise has non-compression nature). This is only metric in VQMT now that can detect increasing of subjective quality in comparison to original.

Sometimes, codec can allow geometry transformation (like shift of heterogeneous objects in frame), that not critical for subjective perception. Objective-reference metrics are very perceptive to such transformation, and in this cases no-reference metric can show result closer to subjective score.

Benchmark

Implementation System & settings Resulution FPS Sec. per frame
VQMT 13 OpenCL-Y NVidia GeForce GTX 660 Ti HD 720p 92.981 0.011
VQMT 13 OpenCL-Y NVidia GeForce GTX 660 Ti FullHD 1080p 41.738 0.024
VQMT 13 OpenCL-Y NVidia GeForce GTX 660 Ti UHD 4k 2160p 10.481 0.095
VQMT 13 default-Y 8 cores HD 720p 43.979 0.023
VQMT 13 default-Y 1 core enabled HD 720p 10.670 0.094
VQMT 13 default-Y 8 cores FullHD 1080p 19.645 0.051
VQMT 13 default-Y 1 core enabled FullHD 1080p 4.573 0.219
VQMT 13 default-Y 8 cores UHD 4k 2160p 4.808 0.208
VQMT 13 default-Y 1 core enabled UHD 4k 2160p 1.136 0.880
measurements were done on VQMT 12 PRO for Windows. Can vary depending on system configuration, input format and other factors

MSU VQMT parameters description

Mean threshold
  • Description. Values of metric greater than this value will be skipped during mean calculation. 0 for disable skipping
  • Default value. 15.
  • Usage. -set "mean_thresh=<value>", where <value> can be:
    any floating point number
Threshold smoothing
  • Description. Values of metric greater than 'Mean threshold' + 'Threshold smoothing' will be skipped, values less than 'Mean threshold' - 'Threshold smoothing' will be assumed with weight 1. Intermediate values will be taken with intermediate weight
  • Default value. 5.
  • Usage. -set "mean_thresh_smoothing=<value>", where <value> can be:
    any floating point number
Type of normalization
  • Description. Can be: fast - the fastest algorithm, low precision; native - like in native NIQE implementation. Slowest one; precise - the most precise algorithm
  • Default value. native
  • Usage. -set "norm_alg=<value>", where <value> can be:
    fast, native, precise

Example

Choose example:
Original, NIQE-Y=3.63
LQ H264, NIQE-Y=5.10
MSU VQMT visualization
Original, NIQE-Y=3.63
Blurring, NIQE-Y=9.47
MSU VQMT visualization
Original, NIQE-Y=3.63
Random points, NIQE-Y=3.80
MSU VQMT visualization
Original, NIQE-Y=3.63
Luminance shift, NIQE-Y=3.63
MSU VQMT visualization
Original, NIQE-Y=3.63
JPEG Q=2, NIQE-Y=20.37
MSU VQMT visualization
Original, NIQE-Y=3.63
JPEG Q=5, NIQE-Y=11.92
MSU VQMT visualization
Original, NIQE-Y=3.63
JPEG Q=10, NIQE-Y=8.30
MSU VQMT visualization
Original, NIQE-Y=3.63
JPEG Q=15, NIQE-Y=6.88
MSU VQMT visualization
Original, NIQE-Y=3.63
JPEG Q=20, NIQE-Y=6.35
MSU VQMT visualization
Original, NIQE-Y=3.63
JPEG Q=40, NIQE-Y=4.58
MSU VQMT visualization
Original, NIQE-Y=3.63
JPEG Q=80, NIQE-Y=3.55
MSU VQMT visualization
Choose example:
Original, NIQE-Y=3.22
LQ H264, NIQE-Y=8.90
MSU VQMT visualization
Original, NIQE-Y=3.22
Blurring, NIQE-Y=9.19
MSU VQMT visualization
Original, NIQE-Y=3.22
Random points, NIQE-Y=3.27
MSU VQMT visualization
Original, NIQE-Y=3.22
Luminance shift, NIQE-Y=3.22
MSU VQMT visualization
Original, NIQE-Y=3.22
JPEG Q=2, NIQE-Y=17.10
MSU VQMT visualization
Original, NIQE-Y=3.22
JPEG Q=5, NIQE-Y=11.14
MSU VQMT visualization
Original, NIQE-Y=3.22
JPEG Q=10, NIQE-Y=6.85
MSU VQMT visualization
Original, NIQE-Y=3.22
JPEG Q=15, NIQE-Y=5.57
MSU VQMT visualization
Original, NIQE-Y=3.22
JPEG Q=20, NIQE-Y=4.95
MSU VQMT visualization
Original, NIQE-Y=3.22
JPEG Q=40, NIQE-Y=3.74
MSU VQMT visualization
Original, NIQE-Y=3.22
JPEG Q=80, NIQE-Y=2.76
MSU VQMT visualization
Choose example:
Original, NIQE-Y=5.16
LQ H264, NIQE-Y=6.00
MSU VQMT visualization
Original, NIQE-Y=5.16
Blurring, NIQE-Y=8.20
MSU VQMT visualization
Original, NIQE-Y=5.16
Random points, NIQE-Y=5.42
MSU VQMT visualization
Original, NIQE-Y=5.16
Luminance shift, NIQE-Y=5.22
MSU VQMT visualization
Original, NIQE-Y=5.16
JPEG Q=2, NIQE-Y=12.31
MSU VQMT visualization
Original, NIQE-Y=5.16
JPEG Q=5, NIQE-Y=9.89
MSU VQMT visualization
Original, NIQE-Y=5.16
JPEG Q=10, NIQE-Y=7.80
MSU VQMT visualization
Original, NIQE-Y=5.16
JPEG Q=15, NIQE-Y=6.87
MSU VQMT visualization
Original, NIQE-Y=5.16
JPEG Q=20, NIQE-Y=6.70
MSU VQMT visualization
Original, NIQE-Y=5.16
JPEG Q=40, NIQE-Y=5.79
MSU VQMT visualization
Original, NIQE-Y=5.16
JPEG Q=80, NIQE-Y=4.91
MSU VQMT visualization

SSIM-family

Structural Similarity (SSIM)

General info

Metric type:
full-reference image metric
Value range:
(images are different) -1..1 (images are same)
Value interpretation:
bigger is better quality
MSU VQMT implementations:
CPU multithreaded fast (default),
cpu multithreaded precice,
cpu multithreaded GPU identical,
OpenCL (recommended),
CUDA
MSU VQMT visualization:
pixel-wise
Available colorspaces:
Y, U, V
Output values:
metric value
Aggregated values:
standard set
MSU VQMT usages:
-metr ssim [over <color components>]
-metr ssim_precise [over <color components>]
-metr ssim_gpu_id [over <color components>]
-metr ssim_cuda [over <color components>]
-metr ssim [over <color components>] -dev <OpenCL device>
External links:
original paper (Z. Wang, A. C. Bovik, H. R. Sheikh and E. P. Simoncelli)

Algorithm description

Main idea of the structure similarity index (SSIM) is to compare distortion of three image components:

  • Luminance comparison
  • Contrast comparison
  • Structure comparison

This algorithm uses some window function , and performs convolution with this window, defined as follows:

In fast implementation, it uses box window: , in other implementations (precise, CUDA, OpenCL, GPU identical) it uses Gaussian window with σ= 1.5, N=10. You can note, that formula above uses negative indexes, and indexes out of image area. We should define image values outside of it's edge. In VQMT we spread closest edge pixel to the desired position.

SSIM uses the following convolutions:

And the computed SSIM in each pixel by the following formula:

where , . Destination metric value is arithmetic mean of SSIM values for each pixel. You also can see SSIM for each individual pixel on visualization.

GPU identical, CUDA, OpenCL implementations should produce very similar result. They use Gaussian window as presice implementation. Value of Presice implementation can differs from these ones.

Benchmark

Implementation System & settings Resulution FPS Sec. per frame
VQMT 13 OpenCL-Y NVidia GeForce GTX 660 Ti HD 720p 356.442 0.003
VQMT 13 OpenCL-Y NVidia GeForce GTX 660 Ti FullHD 1080p 162.801 0.006
VQMT 13 OpenCL-Y NVidia GeForce GTX 660 Ti UHD 4k 2160p 42.519 0.024
VQMT 13 Fast-Y 8 cores HD 720p 228.253 0.004
VQMT 13 Fast-Y 1 core enabled HD 720p 80.533 0.012
VQMT 13 Fast-Y 8 cores FullHD 1080p 99.314 0.010
VQMT 13 Fast-Y 1 core enabled FullHD 1080p 36.313 0.028
VQMT 13 Fast-Y 8 cores UHD 4k 2160p 21.397 0.047
VQMT 13 Fast-Y 1 core enabled UHD 4k 2160p 9.369 0.107
VQMT 13 Precise-Y 8 cores HD 720p 100.199 0.010
VQMT 13 Precise-Y 1 core enabled HD 720p 49.398 0.020
VQMT 13 Precise-Y 8 cores FullHD 1080p 44.808 0.022
VQMT 13 Precise-Y 1 core enabled FullHD 1080p 22.163 0.045
VQMT 13 Precise-Y 8 cores UHD 4k 2160p 10.704 0.093
VQMT 13 Precise-Y 1 core enabled UHD 4k 2160p 5.640 0.177
measurements were done on VQMT 12 PRO for Windows. Can vary depending on system configuration, input format and other factors

Example (SSIM Precise)

Choose example:
Original
LQ H264, SSIM-Y=0.8847
MSU VQMT visualization
Original
Blurring, SSIM-Y=0.8663
MSU VQMT visualization
Original
Random points, SSIM-Y=0.9819
MSU VQMT visualization
Original
Luminance shift, SSIM-Y=0.9995
MSU VQMT visualization
Original
JPEG Q=2, SSIM-Y=0.7502
MSU VQMT visualization
Original
JPEG Q=5, SSIM-Y=0.8035
MSU VQMT visualization
Original
JPEG Q=10, SSIM-Y=0.8749
MSU VQMT visualization
Original
JPEG Q=15, SSIM-Y=0.9073
MSU VQMT visualization
Original
JPEG Q=20, SSIM-Y=0.9265
MSU VQMT visualization
Original
JPEG Q=40, SSIM-Y=0.9584
MSU VQMT visualization
Original
JPEG Q=80, SSIM-Y=0.9830
MSU VQMT visualization
Choose example:
Original
LQ H264, SSIM-Y=0.5180
MSU VQMT visualization
Original
Blurring, SSIM-Y=0.7154
MSU VQMT visualization
Original
Random points, SSIM-Y=0.9867
MSU VQMT visualization
Original
Luminance shift, SSIM-Y=0.9802
MSU VQMT visualization
Original
JPEG Q=2, SSIM-Y=0.5020
MSU VQMT visualization
Original
JPEG Q=5, SSIM-Y=0.6646
MSU VQMT visualization
Original
JPEG Q=10, SSIM-Y=0.7876
MSU VQMT visualization
Original
JPEG Q=15, SSIM-Y=0.8434
MSU VQMT visualization
Original
JPEG Q=20, SSIM-Y=0.8736
MSU VQMT visualization
Original
JPEG Q=40, SSIM-Y=0.9275
MSU VQMT visualization
Original
JPEG Q=80, SSIM-Y=0.9712
MSU VQMT visualization
Choose example:
Original
LQ H264, SSIM-Y=0.9297
MSU VQMT visualization
Original
Blurring, SSIM-Y=0.8839
MSU VQMT visualization
Original
Random points, SSIM-Y=0.9564
MSU VQMT visualization
Original
Luminance shift, SSIM-Y=0.9035
MSU VQMT visualization
Original
JPEG Q=2, SSIM-Y=0.7589
MSU VQMT visualization
Original
JPEG Q=5, SSIM-Y=0.6296
MSU VQMT visualization
Original
JPEG Q=10, SSIM-Y=0.9028
MSU VQMT visualization
Original
JPEG Q=15, SSIM-Y=0.8341
MSU VQMT visualization
Original
JPEG Q=20, SSIM-Y=0.9351
MSU VQMT visualization
Original
JPEG Q=40, SSIM-Y=0.9222
MSU VQMT visualization
Original
JPEG Q=80, SSIM-Y=0.9653
MSU VQMT visualization

Example (SSIM Fast)

Choose example:
Original
LQ H264, SSIM fast-Y=0.8768
MSU VQMT visualization
Original
Blurring, SSIM fast-Y=0.8613
MSU VQMT visualization
Original
Random points, SSIM fast-Y=0.9791
MSU VQMT visualization
Original
Luminance shift, SSIM fast-Y=0.9994
MSU VQMT visualization
Original
JPEG Q=2, SSIM fast-Y=0.7188
MSU VQMT visualization
Original
JPEG Q=5, SSIM fast-Y=0.7839
MSU VQMT visualization
Original
JPEG Q=10, SSIM fast-Y=0.8701
MSU VQMT visualization
Original
JPEG Q=15, SSIM fast-Y=0.9071
MSU VQMT visualization
Original
JPEG Q=20, SSIM fast-Y=0.9283
MSU VQMT visualization
Original
JPEG Q=40, SSIM fast-Y=0.9614
MSU VQMT visualization
Original
JPEG Q=80, SSIM fast-Y=0.9848
MSU VQMT visualization
Choose example:
Original
LQ H264, SSIM fast-Y=0.4736
MSU VQMT visualization
Original
Blurring, SSIM fast-Y=0.7195
MSU VQMT visualization
Original
Random points, SSIM fast-Y=0.9859
MSU VQMT visualization
Original
Luminance shift, SSIM fast-Y=0.9822
MSU VQMT visualization
Original
JPEG Q=2, SSIM fast-Y=0.4943
MSU VQMT visualization
Original
JPEG Q=5, SSIM fast-Y=0.6686
MSU VQMT visualization
Original
JPEG Q=10, SSIM fast-Y=0.8035
MSU VQMT visualization
Original
JPEG Q=15, SSIM fast-Y=0.8597
MSU VQMT visualization
Original
JPEG Q=20, SSIM fast-Y=0.8892
MSU VQMT visualization
Original
JPEG Q=40, SSIM fast-Y=0.9387
MSU VQMT visualization
Original
JPEG Q=80, SSIM fast-Y=0.9764
MSU VQMT visualization
Choose example:
Original
LQ H264, SSIM fast-Y=0.9242
MSU VQMT visualization
Original
Blurring, SSIM fast-Y=0.8806
MSU VQMT visualization
Original
Random points, SSIM fast-Y=0.9531
MSU VQMT visualization
Original
Luminance shift, SSIM fast-Y=0.9075
MSU VQMT visualization
Original
JPEG Q=2, SSIM fast-Y=0.7479
MSU VQMT visualization
Original
JPEG Q=5, SSIM fast-Y=0.6298
MSU VQMT visualization
Original
JPEG Q=10, SSIM fast-Y=0.9004
MSU VQMT visualization
Original
JPEG Q=15, SSIM fast-Y=0.8365
MSU VQMT visualization
Original
JPEG Q=20, SSIM fast-Y=0.9347
MSU VQMT visualization
Original
JPEG Q=40, SSIM fast-Y=0.9230
MSU VQMT visualization
Original
JPEG Q=80, SSIM fast-Y=0.9642
MSU VQMT visualization

Multi-Scale Structural Similarity (MSSSIM)

General info

Metric type:
full-reference image metric
Value range:
(images are different) -1..1 (images are same)
Value interpretation:
bigger is better quality
MSU VQMT implementations:
CPU multithreaded fast (default),
cpu multithreaded precice,
cpu multithreaded GPU identical,
OpenCL (recommended),
CUDA
MSU VQMT visualization:
pixel-wise
Available colorspaces:
Y, U, V
Output values:
metric value
Aggregated values:
standard set
MSU VQMT usages:
-metr msssim [over <color components>]
-metr msssim_precise [over <color components>]
-metr msssim_gpu_id [over <color components>]
-metr msssim_cuda [over <color components>]
-metr msssim [over <color components>] -dev <OpenCL device>
Other names:
MS-SSIM
External links:
original paper (Z. Wang, A. C. Bovik and E. P. Simoncelli)

Algorithm description

This metric performs SSIM calculation as described in SSIM paragraph for 5 scales of input images. Each next scale divides width and height by 2. The result SSIM values are producted with the following powers: 0.0448, 0.2856, 0.3001, 0.2363, 0.1333.

Benchmark

Implementation System & settings Resulution FPS Sec. per frame
VQMT 13 OpenCL-Y NVidia GeForce GTX 660 Ti HD 720p 292.609 0.003
VQMT 13 OpenCL-Y NVidia GeForce GTX 660 Ti FullHD 1080p 153.978 0.006
VQMT 13 OpenCL-Y NVidia GeForce GTX 660 Ti UHD 4k 2160p 41.315 0.024
VQMT 13 Fast-Y 8 cores HD 720p 140.919 0.007
VQMT 13 Fast-Y 1 core enabled HD 720p 59.057 0.017
VQMT 13 Fast-Y 8 cores FullHD 1080p 62.080 0.016
VQMT 13 Fast-Y 1 core enabled FullHD 1080p 26.435 0.038
VQMT 13 Fast-Y 8 cores UHD 4k 2160p 13.583 0.074
VQMT 13 Fast-Y 1 core enabled UHD 4k 2160p 6.681 0.150
VQMT 13 Precise-Y 8 cores HD 720p 60.840 0.016
VQMT 13 Precise-Y 1 core enabled HD 720p 28.119 0.036
VQMT 13 Precise-Y 8 cores FullHD 1080p 27.262 0.037
VQMT 13 Precise-Y 1 core enabled FullHD 1080p 12.176 0.082
VQMT 13 Precise-Y 8 cores UHD 4k 2160p 6.669 0.150
VQMT 13 Precise-Y 1 core enabled UHD 4k 2160p 3.132 0.319
measurements were done on VQMT 12 PRO for Windows. Can vary depending on system configuration, input format and other factors

Example (MSSSIM Precise)

Choose example:
Original
LQ H264, MSSSIM-Y=0.9371
MSU VQMT visualization
Original
Blurring, MSSSIM-Y=0.9450
MSU VQMT visualization
Original
Random points, MSSSIM-Y=0.9925
MSU VQMT visualization
Original
Luminance shift, MSSSIM-Y=0.9994
MSU VQMT visualization
Original
JPEG Q=2, MSSSIM-Y=0.7270
MSU VQMT visualization
Original
JPEG Q=5, MSSSIM-Y=0.8206
MSU VQMT visualization
Original
JPEG Q=10, MSSSIM-Y=0.9264
MSU VQMT visualization
Original
JPEG Q=15, MSSSIM-Y=0.9547
MSU VQMT visualization
Original
JPEG Q=20, MSSSIM-Y=0.9697
MSU VQMT visualization
Original
JPEG Q=40, MSSSIM-Y=0.9885
MSU VQMT visualization
Original
JPEG Q=80, MSSSIM-Y=0.9971
MSU VQMT visualization
Choose example:
Original
LQ H264, MSSSIM-Y=0.5262
MSU VQMT visualization
Original
Blurring, MSSSIM-Y=0.8893
MSU VQMT visualization
Original
Random points, MSSSIM-Y=0.9941
MSU VQMT visualization
Original
Luminance shift, MSSSIM-Y=0.9941
MSU VQMT visualization
Original
JPEG Q=2, MSSSIM-Y=0.6993
MSU VQMT visualization
Original
JPEG Q=5, MSSSIM-Y=0.8299
MSU VQMT visualization
Original
JPEG Q=10, MSSSIM-Y=0.9245
MSU VQMT visualization
Original
JPEG Q=15, MSSSIM-Y=0.9556
MSU VQMT visualization
Original
JPEG Q=20, MSSSIM-Y=0.9690
MSU VQMT visualization
Original
JPEG Q=40, MSSSIM-Y=0.9873
MSU VQMT visualization
Original
JPEG Q=80, MSSSIM-Y=0.9967
MSU VQMT visualization
Choose example:
Original
LQ H264, MSSSIM-Y=0.9608
MSU VQMT visualization
Original
Blurring, MSSSIM-Y=0.9108
MSU VQMT visualization
Original
Random points, MSSSIM-Y=0.9408
MSU VQMT visualization
Original
Luminance shift, MSSSIM-Y=0.9678
MSU VQMT visualization
Original
JPEG Q=2, MSSSIM-Y=0.8103
MSU VQMT visualization
Original
JPEG Q=5, MSSSIM-Y=0.7607
MSU VQMT visualization
Original
JPEG Q=10, MSSSIM-Y=0.9168
MSU VQMT visualization
Original
JPEG Q=15, MSSSIM-Y=0.9018
MSU VQMT visualization
Original
JPEG Q=20, MSSSIM-Y=0.9359
MSU VQMT visualization
Original
JPEG Q=40, MSSSIM-Y=0.9356
MSU VQMT visualization
Original
JPEG Q=80, MSSSIM-Y=0.9473
MSU VQMT visualization

Example (MSSSIM Fast)

Choose example:
Original
LQ H264, MSSSIM fast-Y=0.9125
MSU VQMT visualization
Original
Blurring, MSSSIM fast-Y=0.9181
MSU VQMT visualization
Original
Random points, MSSSIM fast-Y=0.9869
MSU VQMT visualization
Original
Luminance shift, MSSSIM fast-Y=0.9996
MSU VQMT visualization
Original
JPEG Q=2, MSSSIM fast-Y=0.6816
MSU VQMT visualization
Original
JPEG Q=5, MSSSIM fast-Y=0.7807
MSU VQMT visualization
Original
JPEG Q=10, MSSSIM fast-Y=0.9023
MSU VQMT visualization
Original
JPEG Q=15, MSSSIM fast-Y=0.9364
MSU VQMT visualization
Original
JPEG Q=20, MSSSIM fast-Y=0.9554
MSU VQMT visualization
Original
JPEG Q=40, MSSSIM fast-Y=0.9803
MSU VQMT visualization
Original
JPEG Q=80, MSSSIM fast-Y=0.9937
MSU VQMT visualization
Choose example:
Original
LQ H264, MSSSIM fast-Y=0.4858
MSU VQMT visualization
Original
Blurring, MSSSIM fast-Y=0.8364
MSU VQMT visualization
Original
Random points, MSSSIM fast-Y=0.9906
MSU VQMT visualization
Original
Luminance shift, MSSSIM fast-Y=0.9958
MSU VQMT visualization
Original
JPEG Q=2, MSSSIM fast-Y=0.6550
MSU VQMT visualization
Original
JPEG Q=5, MSSSIM fast-Y=0.7865
MSU VQMT visualization
Original
JPEG Q=10, MSSSIM fast-Y=0.8919
MSU VQMT visualization
Original
JPEG Q=15, MSSSIM fast-Y=0.9305
MSU VQMT visualization
Original
JPEG Q=20, MSSSIM fast-Y=0.9485
MSU VQMT visualization
Original
JPEG Q=40, MSSSIM fast-Y=0.9750
MSU VQMT visualization
Original
JPEG Q=80, MSSSIM fast-Y=0.9916
MSU VQMT visualization
Choose example:
Original
LQ H264, MSSSIM fast-Y=0.9476
MSU VQMT visualization
Original
Blurring, MSSSIM fast-Y=0.8916
MSU VQMT visualization
Original
Random points, MSSSIM fast-Y=0.9336
MSU VQMT visualization
Original
Luminance shift, MSSSIM fast-Y=0.9989
MSU VQMT visualization
Original
JPEG Q=2, MSSSIM fast-Y=0.7885
MSU VQMT visualization
Original
JPEG Q=5, MSSSIM fast-Y=0.8424
MSU VQMT visualization
Original
JPEG Q=10, MSSSIM fast-Y=0.9021
MSU VQMT visualization
Original
JPEG Q=15, MSSSIM fast-Y=0.9182
MSU VQMT visualization
Original
JPEG Q=20, MSSSIM fast-Y=0.9263
MSU VQMT visualization
Original
JPEG Q=40, MSSSIM fast-Y=0.9371
MSU VQMT visualization
Original
JPEG Q=80, MSSSIM fast-Y=0.9425
MSU VQMT visualization

Three-commponent Structural Similarity (3SSIM)

General info

Metric type:
full-reference image metric
Value range:
(images are different) -1..1 (images are same)
Value interpretation:
bigger is better quality
MSU VQMT implementations:
CPU multithreaded (default),
OpenCL (recommended),
CUDA
Available colorspaces:
Y, U, V
Output values:
metric value
Aggregated values:
standard set
MSU VQMT usages:
-metr 3ssim [over <color components>]
-metr 3ssim_cuda [over <color components>]
-metr 3ssim [over <color components>] -dev <OpenCL device>
Other names:
3-SSIM
External links:
original paper (C. Li and A. C. Bovik)

Algorithm description

3-Component SSIM Index based on region division of source frames. There are 3 types of regions - edges, textures and smooth regions. Result metric calculated as weighted average of SSIM metric for those regions. In fact, human eye can see difference more precisely on textured or edge regions than on smooth regions. Division based on gradient magnitude is presented in every pixel of images.

Benchmark

Implementation System & settings Resulution FPS Sec. per frame
VQMT 13 OpenCL-Y NVidia GeForce GTX 660 Ti HD 720p 81.189 0.012
VQMT 13 OpenCL-Y NVidia GeForce GTX 660 Ti FullHD 1080p 86.813 0.012
VQMT 13 OpenCL-Y NVidia GeForce GTX 660 Ti UHD 4k 2160p 9.364 0.107
VQMT 13 default-Y 8 cores HD 720p 52.110 0.019
VQMT 13 default-Y 1 core enabled HD 720p 30.669 0.033
VQMT 13 default-Y 8 cores FullHD 1080p 23.131 0.043
VQMT 13 default-Y 1 core enabled FullHD 1080p 13.018 0.077
VQMT 13 default-Y 8 cores UHD 4k 2160p 5.698 0.176
VQMT 13 default-Y 1 core enabled UHD 4k 2160p 3.350 0.299
measurements were done on VQMT 12 PRO for Windows. Can vary depending on system configuration, input format and other factors

Example

Choose example:
Original
LQ H264, 3SSIM-Y=0.7319
MSU VQMT visualization
Original
Blurring, 3SSIM-Y=0.5466
MSU VQMT visualization
Original
Random points, 3SSIM-Y=0.9618
MSU VQMT visualization
Original
Luminance shift, 3SSIM-Y=0.9999
MSU VQMT visualization
Original
JPEG Q=2, 3SSIM-Y=0.4613
MSU VQMT visualization
Original
JPEG Q=5, 3SSIM-Y=0.6302
MSU VQMT visualization
Original
JPEG Q=10, 3SSIM-Y=0.7925
MSU VQMT visualization
Original
JPEG Q=15, 3SSIM-Y=0.8525
MSU VQMT visualization
Original
JPEG Q=20, 3SSIM-Y=0.8856
MSU VQMT visualization
Original
JPEG Q=40, 3SSIM-Y=0.9378
MSU VQMT visualization
Original
JPEG Q=80, 3SSIM-Y=0.9788
MSU VQMT visualization
Choose example:
Original
LQ H264, 3SSIM-Y=0.2680
MSU VQMT visualization
Original
Blurring, 3SSIM-Y=0.4562
MSU VQMT visualization
Original
Random points, 3SSIM-Y=0.9819
MSU VQMT visualization
Original
Luminance shift, 3SSIM-Y=0.9960
MSU VQMT visualization
Original
JPEG Q=2, 3SSIM-Y=0.4636
MSU VQMT visualization
Original
JPEG Q=5, 3SSIM-Y=0.6541
MSU VQMT visualization
Original
JPEG Q=10, 3SSIM-Y=0.8013
MSU VQMT visualization
Original
JPEG Q=15, 3SSIM-Y=0.8592
MSU VQMT visualization
Original
JPEG Q=20, 3SSIM-Y=0.8878
MSU VQMT visualization
Original
JPEG Q=40, 3SSIM-Y=0.9366
MSU VQMT visualization
Original
JPEG Q=80, 3SSIM-Y=0.9778
MSU VQMT visualization
Choose example:
Original
LQ H264, 3SSIM-Y=0.9139
MSU VQMT visualization
Original
Blurring, 3SSIM-Y=0.6577
MSU VQMT visualization
Original
Random points, 3SSIM-Y=0.9556
MSU VQMT visualization
Original
Luminance shift, 3SSIM-Y=0.9835
MSU VQMT visualization
Original
JPEG Q=2, 3SSIM-Y=0.6735
MSU VQMT visualization
Original
JPEG Q=5, 3SSIM-Y=0.8114
MSU VQMT visualization
Original
JPEG Q=10, 3SSIM-Y=0.9235
MSU VQMT visualization
Original
JPEG Q=15, 3SSIM-Y=0.9298
MSU VQMT visualization
Original
JPEG Q=20, 3SSIM-Y=0.9544
MSU VQMT visualization
Original
JPEG Q=40, 3SSIM-Y=0.9630
MSU VQMT visualization
Original
JPEG Q=80, 3SSIM-Y=0.9765
MSU VQMT visualization

Spatio-Temporal SSIM Index

General info

Metric type:
full-reference temporal metric
MSU VQMT implementations:
this metric was temporary excluded from VQMT due to unstability of results
External links:
original paper (A. K. Moorthy and A. C. Bovik)

Algorithm description

The idea of this algorithm is to use motion-oriented weighted windows for SSIM Index. MSU Motion Estimation algorithm is used to retrieve this information. Based on the ME results, weighting window is constructed for every pixel. This window can use up to 33 consecutive frames (16 + current frame + 16). Then SSIM Index is calculated for every window to take into account temporal distortions as well. In addition, another spooling technique is used in this implementation. We use only lower 6% of metric values for the frame to calculate frame metric value. This causes larger metric values difference for difference files.

Legacy notes

This metric was temporary removed in VQMT 10.

Norm calculation metrics

Identity

General info

Metric type:
full-reference image metric
Value range:
(images are different) 0..1 (images are same)
Value interpretation:
binary mode: 1 - images are same, 0 - images are not same;
pixel mode: proportion of same pixels
MSU VQMT implementations:
CPU multithreaded
MSU VQMT visualization:
pixel-wise
Available colorspaces:
R, G, B, Y, U, V, L, RGB, YUV
Output values:
metric value
Aggregated values:
standard set
MSU VQMT usages:
-metr identity [over <color components>]
Other names:
inverse L∞-norm

Algorithm description

Metric cas to modes: binary and pixels. In binary mode only two values are possible: 1 if images are pixel-wise similar, 0 if images have at least 1 different value pixel.

In pixel mode the value is proporsion of similar pixels. 1 means all pixels are similar, 0 means all pixels are dirrefent.

MSU VQMT parameters description

Mode
  • Description. In binary mode only two values are possible: 1 if images are pixel-wise similar, 0 if images have at least 1 different value pixel. In pixel mode the value is proporsion of similar pixels. 1 means all pixels are similar, 0 means all pixels are dirrefent.
  • Default value. binary
  • Usage. -set "mode=<value>", where <value> can be:
    binary, pixels

Benchmark

Implementation System & settings Resulution FPS Sec. per frame
VQMT 13 default-Y 8 cores HD 720p 476.267 0.002
VQMT 13 default-Y 1 core enabled HD 720p 517.130 0.002
VQMT 13 default-Y 8 cores FullHD 1080p 236.593 0.004
VQMT 13 default-Y 1 core enabled FullHD 1080p 236.156 0.004
VQMT 13 default-Y 8 cores UHD 4k 2160p 53.023 0.019
VQMT 13 default-Y 1 core enabled UHD 4k 2160p 53.498 0.019
VQMT 13 default-YUV 8 cores HD 720p 425.065 0.002
VQMT 13 default-YUV 1 core enabled HD 720p 328.717 0.003
VQMT 13 default-YUV 8 cores FullHD 1080p 189.175 0.005
VQMT 13 default-YUV 1 core enabled FullHD 1080p 161.241 0.006
VQMT 13 default-YUV 8 cores UHD 4k 2160p 42.603 0.023
VQMT 13 default-YUV 1 core enabled UHD 4k 2160p 39.692 0.025
measurements were done on VQMT 12 PRO for Windows. Can vary depending on system configuration, input format and other factors

Example

Choose example:
Original
LQ H264, Identity-YUV=0.0000
MSU VQMT visualization
Original
Blurring, Identity-YUV=0.0000
MSU VQMT visualization
Original
Random points, Identity-YUV=0.0000
MSU VQMT visualization
Original
Luminance shift, Identity-YUV=0.0000
MSU VQMT visualization
Original
JPEG Q=2, Identity-YUV=0.0000
MSU VQMT visualization
Original
JPEG Q=5, Identity-YUV=0.0000
MSU VQMT visualization
Original
JPEG Q=10, Identity-YUV=0.0000
MSU VQMT visualization
Original
JPEG Q=15, Identity-YUV=0.0000
MSU VQMT visualization
Original
JPEG Q=20, Identity-YUV=0.0000
MSU VQMT visualization
Original
JPEG Q=40, Identity-YUV=0.0000
MSU VQMT visualization
Original
JPEG Q=80, Identity-YUV=0.0000
MSU VQMT visualization
Choose example:
Original
LQ H264, Identity-YUV=0.0000
MSU VQMT visualization
Original
Blurring, Identity-YUV=0.0000
MSU VQMT visualization
Original
Random points, Identity-YUV=0.0000
MSU VQMT visualization
Original
Luminance shift, Identity-YUV=0.0000
MSU VQMT visualization
Original
JPEG Q=2, Identity-YUV=0.0000
MSU VQMT visualization
Original
JPEG Q=5, Identity-YUV=0.0000
MSU VQMT visualization
Original
JPEG Q=10, Identity-YUV=0.0000
MSU VQMT visualization
Original
JPEG Q=15, Identity-YUV=0.0000
MSU VQMT visualization
Original
JPEG Q=20, Identity-YUV=0.0000
MSU VQMT visualization
Original
JPEG Q=40, Identity-YUV=0.0000
MSU VQMT visualization
Original
JPEG Q=80, Identity-YUV=0.0000
MSU VQMT visualization
Choose example:
Original
LQ H264, Identity-YUV=0.0000
MSU VQMT visualization
Original
Blurring, Identity-YUV=0.0000
MSU VQMT visualization
Original
Random points, Identity-YUV=0.0000
MSU VQMT visualization
Original
Luminance shift, Identity-YUV=0.0000
MSU VQMT visualization
Original
JPEG Q=2, Identity-YUV=0.0000
MSU VQMT visualization
Original
JPEG Q=5, Identity-YUV=0.0000
MSU VQMT visualization
Original
JPEG Q=10, Identity-YUV=0.0000
MSU VQMT visualization
Original
JPEG Q=15, Identity-YUV=0.0000
MSU VQMT visualization
Original
JPEG Q=20, Identity-YUV=0.0000
MSU VQMT visualization
Original
JPEG Q=40, Identity-YUV=0.0000
MSU VQMT visualization
Original
JPEG Q=80, Identity-YUV=0.0000
MSU VQMT visualization

Legacy notes

This metric was introduce in VQMT 13.

Delta

General info

Metric type:
full-reference image metric
Value range:
(original is much brigther) -1..1 (distorted is much brigther)
Value interpretation:
bigger is darker original, 0 - same brightness
MSU VQMT implementations:
CPU multithreaded
MSU VQMT visualization:
pixel-wise
Available colorspaces:
R, G, B, Y, U, V, L
Output values:
metric value
Aggregated values:
standard set
MSU VQMT usages:
-metr delta [over <color components>]
Other names:
mean difference

Algorithm description

The value of this metric is the mean difference of the color value in the corresponding points of image.

where - video width, - video height, image data are in range 0..1. This formula can be rewriten to the following way: .

This metric doesn't show a quality loss, because it can be 0 for completely different images, but you can detect general brightness shifts using this metric if you are sure images have same structure.

Benchmark

Implementation System & settings Resulution FPS Sec. per frame
VQMT 13 default-Y 8 cores HD 720p 512.883 0.002
VQMT 13 default-Y 1 core enabled HD 720p 495.231 0.002
VQMT 13 default-Y 8 cores FullHD 1080p 221.637 0.005
VQMT 13 default-Y 1 core enabled FullHD 1080p 225.761 0.004
VQMT 13 default-Y 8 cores UHD 4k 2160p 50.758 0.020
VQMT 13 default-Y 1 core enabled UHD 4k 2160p 51.687 0.019
measurements were done on VQMT 12 PRO for Windows. Can vary depending on system configuration, input format and other factors

Example

Choose example:
Original
LQ H264, Delta-Y=0.0011
MSU VQMT visualization
Original
Blurring, Delta-Y=0.0003
MSU VQMT visualization
Original
Random points, Delta-Y=0.0001
MSU VQMT visualization
Original
Luminance shift, Delta-Y=0.0078
MSU VQMT visualization
Original
JPEG Q=2, Delta-Y=-0.0032
MSU VQMT visualization
Original
JPEG Q=5, Delta-Y=-0.0015
MSU VQMT visualization
Original
JPEG Q=10, Delta-Y=0.0005
MSU VQMT visualization
Original
JPEG Q=15, Delta-Y=-0.0009
MSU VQMT visualization
Original
JPEG Q=20, Delta-Y=-0.0002
MSU VQMT visualization
Original
JPEG Q=40, Delta-Y=0.0002
MSU VQMT visualization
Original
JPEG Q=80, Delta-Y=0.0003
MSU VQMT visualization
Choose example:
Original
LQ H264, Delta-Y=0.0078
MSU VQMT visualization
Original
Blurring, Delta-Y=0.0001
MSU VQMT visualization
Original
Random points, Delta-Y=0.0001
MSU VQMT visualization
Original
Luminance shift, Delta-Y=0.0076
MSU VQMT visualization
Original
JPEG Q=2, Delta-Y=-0.0031
MSU VQMT visualization
Original
JPEG Q=5, Delta-Y=0.0009
MSU VQMT visualization
Original
JPEG Q=10, Delta-Y=-0.0001
MSU VQMT visualization
Original
JPEG Q=15, Delta-Y=0.0003
MSU VQMT visualization
Original
JPEG Q=20, Delta-Y=0.0001
MSU VQMT visualization
Original
JPEG Q=40, Delta-Y=0.0000
MSU VQMT visualization
Original
JPEG Q=80, Delta-Y=0.0000
MSU VQMT visualization
Choose example:
Original
LQ H264, Delta-Y=-0.0089
MSU VQMT visualization
Original
Blurring, Delta-Y=0.0066
MSU VQMT visualization
Original
Random points, Delta-Y=0.0063
MSU VQMT visualization
Original
Luminance shift, Delta-Y=0.0078
MSU VQMT visualization
Original
JPEG Q=2, Delta-Y=0.0063
MSU VQMT visualization
Original
JPEG Q=5, Delta-Y=0.0154
MSU VQMT visualization
Original
JPEG Q=10, Delta-Y=0.0065
MSU VQMT visualization
Original
JPEG Q=15, Delta-Y=0.0085
MSU VQMT visualization
Original
JPEG Q=20, Delta-Y=0.0065
MSU VQMT visualization
Original
JPEG Q=40, Delta-Y=0.0074
MSU VQMT visualization
Original
JPEG Q=80, Delta-Y=0.0063
MSU VQMT visualization

Legacy notes

Since VQMT 12 metric uses input range 0..1. In legacy mode is assumed input range to be 0..255 and 0..100 for L channel.

Mean Sum of Absolute Differences (MSAD)

General info

Metric type:
full-reference image metric
Value range:
(same images) 0..1 (completely different)
Value interpretation:
smaller is better quality
MSU VQMT implementations:
CPU multithreaded
MSU VQMT visualization:
pixel-wise
Available colorspaces:
R, G, B, Y, U, V, L
Output values:
metric value
Aggregated values:
standard set
MSU VQMT usages:
-metr msad [over <color components>]
Other names:
SAD, -norm
External links:
Wikipedia

Algorithm description

This metric has very similar formula to Delta, but has a modulo around the difference:

where - video width, - video height, image data are in range 0..1. Metric depends only on difference of original and distorted, it is -norm of this difference.

Unlike Delta, this metric will show real difference between images, 0 means completely equivalent images.

Benchmark

Implementation System & settings Resulution FPS Sec. per frame
VQMT 13 default-Y 8 cores HD 720p 511.594 0.002
VQMT 13 default-Y 1 core enabled HD 720p 497.523 0.002
VQMT 13 default-Y 8 cores FullHD 1080p 224.310 0.004
VQMT 13 default-Y 1 core enabled FullHD 1080p 225.746 0.004
VQMT 13 default-Y 8 cores UHD 4k 2160p 50.950 0.020
VQMT 13 default-Y 1 core enabled UHD 4k 2160p 51.451 0.019
measurements were done on VQMT 12 PRO for Windows. Can vary depending on system configuration, input format and other factors

Example

Choose example:
Original
LQ H264, MSAD-Y=0.0144
MSU VQMT visualization
Original
Blurring, MSAD-Y=0.0127
MSU VQMT visualization
Original
Random points, MSAD-Y=0.0003
MSU VQMT visualization
Original
Luminance shift, MSAD-Y=0.0078
MSU VQMT visualization
Original
JPEG Q=2, MSAD-Y=0.0375
MSU VQMT visualization
Original
JPEG Q=5, MSAD-Y=0.0259
MSU VQMT visualization
Original
JPEG Q=10, MSAD-Y=0.0158
MSU VQMT visualization
Original
JPEG Q=15, MSAD-Y=0.0126
MSU VQMT visualization
Original
JPEG Q=20, MSAD-Y=0.0106
MSU VQMT visualization
Original
JPEG Q=40, MSAD-Y=0.0073
MSU VQMT visualization
Original
JPEG Q=80, MSAD-Y=0.0043
MSU VQMT visualization
Choose example:
Original
LQ H264, MSAD-Y=0.0536
MSU VQMT visualization
Original
Blurring, MSAD-Y=0.0268
MSU VQMT visualization
Original
Random points, MSAD-Y=0.0003
MSU VQMT visualization
Original
Luminance shift, MSAD-Y=0.0076
MSU VQMT visualization
Original
JPEG Q=2, MSAD-Y=0.0431
MSU VQMT visualization
Original
JPEG Q=5, MSAD-Y=0.0320
MSU VQMT visualization
Original
JPEG Q=10, MSAD-Y=0.0223
MSU VQMT visualization
Original
JPEG Q=15, MSAD-Y=0.0184
MSU VQMT visualization
Original
JPEG Q=20, MSAD-Y=0.0163
MSU VQMT visualization
Original
JPEG Q=40, MSAD-Y=0.0119
MSU VQMT visualization
Original
JPEG Q=80, MSAD-Y=0.0073
MSU VQMT visualization
Choose example:
Original
LQ H264, MSAD-Y=0.0174
MSU VQMT visualization
Original
Blurring, MSAD-Y=0.0416
MSU VQMT visualization
Original
Random points, MSAD-Y=0.0324
MSU VQMT visualization
Original
Luminance shift, MSAD-Y=0.0078
MSU VQMT visualization
Original
JPEG Q=2, MSAD-Y=0.0431
MSU VQMT visualization
Original
JPEG Q=5, MSAD-Y=0.0455
MSU VQMT visualization
Original
JPEG Q=10, MSAD-Y=0.0345
MSU VQMT visualization
Original
JPEG Q=15, MSAD-Y=0.0356
MSU VQMT visualization
Original
JPEG Q=20, MSAD-Y=0.0335
MSU VQMT visualization
Original
JPEG Q=40, MSAD-Y=0.0336
MSU VQMT visualization
Original
JPEG Q=80, MSAD-Y=0.0323
MSU VQMT visualization

Legacy notes

Since VQMT 12 metric uses input range 0..1. In legacy mode is assumed input range to be 0..255 and 0..100 for L channel.

Mean Squared Error (MSE)

General info

Metric type:
full-reference image metric
Value range:
(same images) 0..1 (completely different)
Value interpretation:
smaller is better quality
MSU VQMT implementations:
CPU multithreaded
MSU VQMT visualization:
pixel-wise
Available colorspaces:
R, G, B, Y, U, V, L
Output values:
metric value
Aggregated values:
standard set
MSU VQMT usages:
-metr mse [over <color components>]
Other names:
-norm
External links:
Wikipedia

Algorithm description

Metric depends only on difference of original and distorted, it is L2-norm of this difference. Metric could be computed using the following formula:

where - video width, - video height, image data are in range 0..1.

Benchmark

Implementation System & settings Resulution FPS Sec. per frame
VQMT 13 default-Y 8 cores HD 720p 511.686 0.002
VQMT 13 default-Y 1 core enabled HD 720p 494.866 0.002
VQMT 13 default-Y 8 cores FullHD 1080p 224.004 0.004
VQMT 13 default-Y 1 core enabled FullHD 1080p 225.775 0.004
VQMT 13 default-Y 8 cores UHD 4k 2160p 51.577 0.019
VQMT 13 default-Y 1 core enabled UHD 4k 2160p 49.594 0.020
measurements were done on VQMT 12 PRO for Windows. Can vary depending on system configuration, input format and other factors

Example

Choose example:
Original
LQ H264, MSE-Y=0.0005
MSU VQMT visualization
Original
Blurring, MSE-Y=0.0006
MSU VQMT visualization
Original
Random points, MSE-Y=0.0002
MSU VQMT visualization
Original
Luminance shift, MSE-Y=0.0001
MSU VQMT visualization
Original
JPEG Q=2, MSE-Y=0.0022
MSU VQMT visualization
Original
JPEG Q=5, MSE-Y=0.0011
MSU VQMT visualization
Original
JPEG Q=10, MSE-Y=0.0005
MSU VQMT visualization
Original
JPEG Q=15, MSE-Y=0.0003
MSU VQMT visualization
Original
JPEG Q=20, MSE-Y=0.0003
MSU VQMT visualization
Original
JPEG Q=40, MSE-Y=0.0001
MSU VQMT visualization
Original
JPEG Q=80, MSE-Y=0.0000
MSU VQMT visualization
Choose example:
Original
LQ H264, MSE-Y=0.0059
MSU VQMT visualization
Original
Blurring, MSE-Y=0.0022
MSU VQMT visualization
Original
Random points, MSE-Y=0.0002
MSU VQMT visualization
Original
Luminance shift, MSE-Y=0.0001
MSU VQMT visualization
Original
JPEG Q=2, MSE-Y=0.0035
MSU VQMT visualization
Original
JPEG Q=5, MSE-Y=0.0020
MSU VQMT visualization
Original
JPEG Q=10, MSE-Y=0.0011
MSU VQMT visualization
Original
JPEG Q=15, MSE-Y=0.0007
MSU VQMT visualization
Original
JPEG Q=20, MSE-Y=0.0006
MSU VQMT visualization
Original
JPEG Q=40, MSE-Y=0.0003
MSU VQMT visualization
Original
JPEG Q=80, MSE-Y=0.0001
MSU VQMT visualization
Choose example:
Original
LQ H264, MSE-Y=0.0007
MSU VQMT visualization
Original
Blurring, MSE-Y=0.0038
MSU VQMT visualization
Original
Random points, MSE-Y=0.0023
MSU VQMT visualization
Original
Luminance shift, MSE-Y=0.0001
MSU VQMT visualization
Original
JPEG Q=2, MSE-Y=0.0040
MSU VQMT visualization
Original
JPEG Q=5, MSE-Y=0.0032
MSU VQMT visualization
Original
JPEG Q=10, MSE-Y=0.0024
MSU VQMT visualization
Original
JPEG Q=15, MSE-Y=0.0023
MSU VQMT visualization
Original
JPEG Q=20, MSE-Y=0.0023
MSU VQMT visualization
Original
JPEG Q=40, MSE-Y=0.0022
MSU VQMT visualization
Original
JPEG Q=80, MSE-Y=0.0021
MSU VQMT visualization

Legacy notes

Since VQMT 12 metric uses input range 0..1. In legacy mode is assumed input range to be 0..255 and 0..100 for L channel.

Other metrics

DCT-based Video Quality Metric (VQM)

General info

Metric type:
full-reference image metric
Value range:
(images are same) 0..∞ (bad quality)
Value interpretation:
smaller is better quality
MSU VQMT implementations:
CPU multithreaded
MSU VQMT visualization:
pixel-wise
Available colorspaces:
Y
Output values:
metric value
Aggregated values:
standard set
MSU VQMT usages:
-metr vqm
External links:
original paper (Feng Xiao)

Algorithm description

This metric uses discrete cosine transform (DCT) to predict human rank. It is different from widely spreaded VQM metric by ITU, that currently not implemented in VQMT. Following calculations are processed to get value of metric:

  • Color transform. YUV color space is used for metric calculation.
  • DCT transform of blocks 8x8. It is used to separate images into different frequencies.
  • Conversion from DCT coefficients to local contrast (LC) using following equation:

where DC is the DCT coefficient with indexes (0, 0).

  • Conversion from LC to just-noticeable difference:

where CSF is Contrast Sensitivity Function. Inverse MPEG-4 default quantization matrix is used as CSF in original article.

  • Weighted pooling of mean and maximum distortions. First, absolute difference D is calculated for JND coefficients following by VQM value construction:

Please, refer the original paper for the details.

Benchmark

Implementation System & settings Resulution FPS Sec. per frame
VQMT 13 default-Y 8 cores HD 720p 103.883 0.010
VQMT 13 default-Y 1 core enabled HD 720p 24.360 0.041
VQMT 13 default-Y 8 cores FullHD 1080p 47.902 0.021
VQMT 13 default-Y 1 core enabled FullHD 1080p 11.205 0.089
VQMT 13 default-Y 8 cores UHD 4k 2160p 12.045 0.083
VQMT 13 default-Y 1 core enabled UHD 4k 2160p 2.848 0.351
measurements were done on VQMT 12 PRO for Windows. Can vary depending on system configuration, input format and other factors

Example

Choose example:
Original
LQ H264, VQM-Y=1.85
MSU VQMT visualization
Original
Blurring, VQM-Y=2.27
MSU VQMT visualization
Original
Random points, VQM-Y=0.5140
MSU VQMT visualization
Original
Luminance shift, VQM-Y=0.4022
MSU VQMT visualization
Original
JPEG Q=2, VQM-Y=3.56
MSU VQMT visualization
Original
JPEG Q=5, VQM-Y=2.45
MSU VQMT visualization
Original
JPEG Q=10, VQM-Y=1.57
MSU VQMT visualization
Original
JPEG Q=15, VQM-Y=1.32
MSU VQMT visualization
Original
JPEG Q=20, VQM-Y=1.16
MSU VQMT visualization
Original
JPEG Q=40, VQM-Y=0.8209
MSU VQMT visualization
Original
JPEG Q=80, VQM-Y=0.5215
MSU VQMT visualization
Choose example:
Original
LQ H264, VQM-Y=8.66
MSU VQMT visualization
Original
Blurring, VQM-Y=4.58
MSU VQMT visualization
Original
Random points, VQM-Y=1.03
MSU VQMT visualization
Original
Luminance shift, VQM-Y=0.9054
MSU VQMT visualization
Original
JPEG Q=2, VQM-Y=6.12
MSU VQMT visualization
Original
JPEG Q=5, VQM-Y=4.81
MSU VQMT visualization
Original
JPEG Q=10, VQM-Y=3.57
MSU VQMT visualization
Original
JPEG Q=15, VQM-Y=3.00
MSU VQMT visualization
Original
JPEG Q=20, VQM-Y=2.81
MSU VQMT visualization
Original
JPEG Q=40, VQM-Y=1.97
MSU VQMT visualization
Original
JPEG Q=80, VQM-Y=1.30
MSU VQMT visualization
Choose example:
Original
LQ H264, VQM-Y=3.08
MSU VQMT visualization
Original
Blurring, VQM-Y=4.49
MSU VQMT visualization
Original
Random points, VQM-Y=3.28
MSU VQMT visualization
Original
Luminance shift, VQM-Y=0.9998
MSU VQMT visualization
Original
JPEG Q=2, VQM-Y=4.97
MSU VQMT visualization
Original
JPEG Q=5, VQM-Y=4.83
MSU VQMT visualization
Original
JPEG Q=10, VQM-Y=3.73
MSU VQMT visualization
Original
JPEG Q=15, VQM-Y=3.77
MSU VQMT visualization
Original
JPEG Q=20, VQM-Y=3.45
MSU VQMT visualization
Original
JPEG Q=40, VQM-Y=3.41
MSU VQMT visualization
Original
JPEG Q=80, VQM-Y=3.14
MSU VQMT visualization

Legacy notes

In VQMT 12 we removed input aligning, also metric is optimized.

MSU Time shift

General info

Metric type:
full-reference temporal metric
Value range:
depending on settings
Value interpretation:
Metric value is shift in frames of distorted sequence relative to original. If n-th value is x, then n+x'th frame of distorted corresponds to n'th frame of original
MSU VQMT implementations:
CPU multithreaded
MSU VQMT visualization:
block-wise
Available colorspaces:
Y
Output values:
metric value
multiple individual measument results if on
Aggregated values:
standard set
MSU VQMT usages:
-metr time-shift

Algorithm description

This metric performes calculation of another metric (base metric: PSNR or SSIM) between each frame of original image and several frames of distorted image. You can choose what interval will be used in settings. This metric can detect such artifacts as skipped frame, duplicated frame, small fps mismatch.

Metric considers, that the best metric value for specific original frame is the correct shift. Metric can prefer smaller shift with base metric value X to bigger shift with metric value Y if , where threshold can be set in metric settings. This helps to avoid random fluctuations.

Also, values of base metric will be smothed over adjacent frames if smoothing is on. This can help in case of very there are very similar adjacent frames in seuqence or the desired frame is absent (for example, negative shift on first frame).

MSU VQMT parameters description

Max. shift
  • Description. Maximum shift, that can be detected. Note: large values leads big memory consumption.
  • Default value. 5.
  • Usage. -set "max-shift=<value>", where <value> can be:
    integer in range 0..25
Direction
  • Description. Detect only positive shifts (frame dups), negatives (frame drops) or both.
  • Default value. both
  • Usage. -set "direction=<value>", where <value> can be:
    positive, negative, both
Destination metric
  • Description. This metric will be used to measure similarity between frames
  • Default value. psnr
  • Usage. -set "metric=<value>", where <value> can be:
    psnr, ssim
Show metric values
  • Description. Metric will output now only shift, but destination metric values
  • Default value. false
  • Usage. -set "show-metric=<value>", where <value> can be:
    true, false
Threshold
  • Description. We will consider shift only if metric for neighbour frame better than this thresold multiplied to metric for similar frame
  • Default value. false
  • Usage. -set "threshold=<value>", where <value> can be:
    any floating point number
Smoothing
  • Description. Will smooth metric values over time. If equal n, than smoothing will be in the interval frame-n..frame+n
  • Default value. 1
  • Usage. -set "smoothing=<value>", where <value> can be:
    integer in range 0..25

Benchmark

Implementation System & settings Resulution FPS Sec. per frame
VQMT 13 default-Y 8 cores HD 720p 341.820 0.003
VQMT 13 default-Y 1 core enabled HD 720p 276.412 0.004
VQMT 13 default-Y 8 cores FullHD 1080p 125.327 0.008
VQMT 13 default-Y 1 core enabled FullHD 1080p 98.154 0.010
VQMT 13 default-Y 8 cores UHD 4k 2160p 30.121 0.033
VQMT 13 default-Y 1 core enabled UHD 4k 2160p 24.025 0.042
measurements were done on VQMT 12 PRO for Windows. Can vary depending on system configuration, input format and other factors

Example

Choose example:
Original
LQ H264, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Original
Blurring, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Original
Random points, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Original
Luminance shift, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Original
JPEG Q=2, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Original
JPEG Q=5, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Original
JPEG Q=10, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Original
JPEG Q=15, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Original
JPEG Q=20, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Original
JPEG Q=40, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Original
JPEG Q=80, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Choose example:
Original
LQ H264, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Original
Blurring, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Original
Random points, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Original
Luminance shift, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Original
JPEG Q=2, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Original
JPEG Q=5, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Original
JPEG Q=10, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Original
JPEG Q=15, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Original
JPEG Q=20, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Original
JPEG Q=40, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Original
JPEG Q=80, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Choose example:
Original
LQ H264, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Original
Blurring, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Original
Random points, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Original
Luminance shift, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Original
JPEG Q=2, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Original
JPEG Q=5, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Original
JPEG Q=10, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Original
JPEG Q=15, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Original
JPEG Q=20, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Original
JPEG Q=40, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization
Original
JPEG Q=80, Time shift (local)_shift-Y=0.0000
MSU VQMT visualization

Legacy notes

This metric was introduced in VQMT 13.

No-reference informational metrics

MSU Blurring Metric

General info

Metric type:
no-reference image metric
Value range:
(constant image) 0..1 (very noisy)
Value interpretation:
bigger is more noise
MSU VQMT implementations:
CPU multithreaded sigma (default),
CPU multithreaded delta
MSU VQMT visualization:
pixel-wise
Available colorspaces:
Y, R, G, B for sigma
Y, U, V, R, G, B for delta
Output values:
metric value
Aggregated values:
standard set
MSU VQMT usages:
-metr blurring [over <color components>]
-metr blurring_delta [over <color components>]

Algorithm description

This metric allows you to compare power of blurring of two images. If value of the metric for first picture is greater than for second, it means that second picture is more blurred, than first.

Main features: this metric is fast and doesn't require source video.

This method estimates color variance in the neighborhood of a pixel and computes average variance. This metric has 2 variations:

  • Sigma (default since VQMT 11). It uses 3-pixel radius neighborhood and normalized Gaussian kernel
  • Delta (the only before VQMT 11). It uses 1-pixel radius neighborhood

Notes:

  • You can't measure blurriness on constant or gradient areas of input image. So, the value of metric is very dependent on amount of edges in images.
  • This metric will detect not only compression artifacts, but natural not-infocus areas, so the value of metric is very dependent on area of focused objects in the frame. You shouldn't use value of this metric as final blurrness index, use is only to compare blurrness of images with similar structure.

Benchmark

Implementation System & settings Resulution FPS Sec. per frame
VQMT 13 Delta-Y 8 cores HD 720p 1062.426 0.001
VQMT 13 Delta-Y 1 core enabled HD 720p 998.102 0.001
VQMT 13 Delta-Y 8 cores FullHD 1080p 451.868 0.002
VQMT 13 Delta-Y 1 core enabled FullHD 1080p 449.640 0.002
VQMT 13 Delta-Y 8 cores UHD 4k 2160p 105.745 0.009
VQMT 13 Delta-Y 1 core enabled UHD 4k 2160p 107.198 0.009
VQMT 13 Sigma-Y 8 cores HD 720p 101.870 0.010
VQMT 13 Sigma-Y 1 core enabled HD 720p 38.839 0.026
VQMT 13 Sigma-Y 8 cores FullHD 1080p 45.462 0.022
VQMT 13 Sigma-Y 1 core enabled FullHD 1080p 17.164 0.058
VQMT 13 Sigma-Y 8 cores UHD 4k 2160p 11.145 0.090
VQMT 13 Sigma-Y 1 core enabled UHD 4k 2160p 4.211 0.237
measurements were done on VQMT 12 PRO for Windows. Can vary depending on system configuration, input format and other factors

Example (Blurring sigma)

Choose example:
Original, Blurring (sigma)-Y=0.0273
LQ H264, Blurring (sigma)-Y=0.0172
MSU VQMT visualization
Original, Blurring (sigma)-Y=0.0273
Blurring, Blurring (sigma)-Y=0.0101
MSU VQMT visualization
Original, Blurring (sigma)-Y=0.0273
Random points, Blurring (sigma)-Y=0.0297
MSU VQMT visualization
Original, Blurring (sigma)-Y=0.0273
Luminance shift, Blurring (sigma)-Y=0.0273
MSU VQMT visualization
Original, Blurring (sigma)-Y=0.0273
JPEG Q=2, Blurring (sigma)-Y=0.0191
MSU VQMT visualization
Original, Blurring (sigma)-Y=0.0273
JPEG Q=5, Blurring (sigma)-Y=0.0222
MSU VQMT visualization
Original, Blurring (sigma)-Y=0.0273
JPEG Q=10, Blurring (sigma)-Y=0.0250
MSU VQMT visualization
Original, Blurring (sigma)-Y=0.0273
JPEG Q=15, Blurring (sigma)-Y=0.0259
MSU VQMT visualization
Original, Blurring (sigma)-Y=0.0273
JPEG Q=20, Blurring (sigma)-Y=0.0263
MSU VQMT visualization
Original, Blurring (sigma)-Y=0.0273
JPEG Q=40, Blurring (sigma)-Y=0.0267
MSU VQMT visualization
Original, Blurring (sigma)-Y=0.0273
JPEG Q=80, Blurring (sigma)-Y=0.0270
MSU VQMT visualization
Choose example:
Original, Blurring (sigma)-Y=0.0567
LQ H264, Blurring (sigma)-Y=0.0119
MSU VQMT visualization
Original, Blurring (sigma)-Y=0.0567
Blurring, Blurring (sigma)-Y=0.0191
MSU VQMT visualization
Original, Blurring (sigma)-Y=0.0567
Random points, Blurring (sigma)-Y=0.0589
MSU VQMT visualization
Original, Blurring (sigma)-Y=0.0567
Luminance shift, Blurring (sigma)-Y=0.0566
MSU VQMT visualization
Original, Blurring (sigma)-Y=0.0567
JPEG Q=2, Blurring (sigma)-Y=0.0403
MSU VQMT visualization
Original, Blurring (sigma)-Y=0.0567
JPEG Q=5, Blurring (sigma)-Y=0.0457
MSU VQMT visualization
Original, Blurring (sigma)-Y=0.0567
JPEG Q=10, Blurring (sigma)-Y=0.0525
MSU VQMT visualization
Original, Blurring (sigma)-Y=0.0567
JPEG Q=15, Blurring (sigma)-Y=0.0541
MSU VQMT visualization
Original, Blurring (sigma)-Y=0.0567
JPEG Q=20, Blurring (sigma)-Y=0.0547
MSU VQMT visualization
Original, Blurring (sigma)-Y=0.0567
JPEG Q=40, Blurring (sigma)-Y=0.0561
MSU VQMT visualization
Original, Blurring (sigma)-Y=0.0567
JPEG Q=80, Blurring (sigma)-Y=0.0569
MSU VQMT visualization
Choose example:
Original, Blurring (sigma)-Y=0.0339
LQ H264, Blurring (sigma)-Y=0.0320
MSU VQMT visualization
Original, Blurring (sigma)-Y=0.0339
Blurring, Blurring (sigma)-Y=0.0200
MSU VQMT visualization
Original, Blurring (sigma)-Y=0.0339
Random points, Blurring (sigma)-Y=0.0374
MSU VQMT visualization
Original, Blurring (sigma)-Y=0.0339
Luminance shift, Blurring (sigma)-Y=0.0339
MSU VQMT visualization
Original, Blurring (sigma)-Y=0.0339
JPEG Q=2, Blurring (sigma)-Y=0.0408
MSU VQMT visualization
Original, Blurring (sigma)-Y=0.0339
JPEG Q=5, Blurring (sigma)-Y=0.0391
MSU VQMT visualization
Original, Blurring (sigma)-Y=0.0339
JPEG Q=10, Blurring (sigma)-Y=0.0376
MSU VQMT visualization
Original, Blurring (sigma)-Y=0.0339
JPEG Q=15, Blurring (sigma)-Y=0.0367
MSU VQMT visualization
Original, Blurring (sigma)-Y=0.0339
JPEG Q=20, Blurring (sigma)-Y=0.0362
MSU VQMT visualization
Original, Blurring (sigma)-Y=0.0339
JPEG Q=40, Blurring (sigma)-Y=0.0353
MSU VQMT visualization
Original, Blurring (sigma)-Y=0.0339
JPEG Q=80, Blurring (sigma)-Y=0.0349
MSU VQMT visualization

Example (Blurring delta)

Choose example:
Original, Blurring (delta)-Y=0.0160
LQ H264, Blurring (delta)-Y=0.0094
MSU VQMT visualization
Original, Blurring (delta)-Y=0.0160
Blurring, Blurring (delta)-Y=0.0053
MSU VQMT visualization
Original, Blurring (delta)-Y=0.0160
Random points, Blurring (delta)-Y=0.0165
MSU VQMT visualization
Original, Blurring (delta)-Y=0.0160
Luminance shift, Blurring (delta)-Y=0.0160
MSU VQMT visualization
Original, Blurring (delta)-Y=0.0160
JPEG Q=2, Blurring (delta)-Y=0.0064
MSU VQMT visualization
Original, Blurring (delta)-Y=0.0160
JPEG Q=5, Blurring (delta)-Y=0.0092
MSU VQMT visualization
Original, Blurring (delta)-Y=0.0160
JPEG Q=10, Blurring (delta)-Y=0.0124
MSU VQMT visualization
Original, Blurring (delta)-Y=0.0160
JPEG Q=15, Blurring (delta)-Y=0.0137
MSU VQMT visualization
Original, Blurring (delta)-Y=0.0160
JPEG Q=20, Blurring (delta)-Y=0.0143
MSU VQMT visualization
Original, Blurring (delta)-Y=0.0160
JPEG Q=40, Blurring (delta)-Y=0.0152
MSU VQMT visualization
Original, Blurring (delta)-Y=0.0160
JPEG Q=80, Blurring (delta)-Y=0.0157
MSU VQMT visualization
Choose example:
Original, Blurring (delta)-Y=0.0313
LQ H264, Blurring (delta)-Y=0.0055
MSU VQMT visualization
Original, Blurring (delta)-Y=0.0313
Blurring, Blurring (delta)-Y=0.0102
MSU VQMT visualization
Original, Blurring (delta)-Y=0.0313
Random points, Blurring (delta)-Y=0.0318
MSU VQMT visualization
Original, Blurring (delta)-Y=0.0313
Luminance shift, Blurring (delta)-Y=0.0313
MSU VQMT visualization
Original, Blurring (delta)-Y=0.0313
JPEG Q=2, Blurring (delta)-Y=0.0153
MSU VQMT visualization
Original, Blurring (delta)-Y=0.0313
JPEG Q=5, Blurring (delta)-Y=0.0211
MSU VQMT visualization
Original, Blurring (delta)-Y=0.0313
JPEG Q=10, Blurring (delta)-Y=0.0271
MSU VQMT visualization
Original, Blurring (delta)-Y=0.0313
JPEG Q=15, Blurring (delta)-Y=0.0291
MSU VQMT visualization
Original, Blurring (delta)-Y=0.0313
JPEG Q=20, Blurring (delta)-Y=0.0299
MSU VQMT visualization
Original, Blurring (delta)-Y=0.0313
JPEG Q=40, Blurring (delta)-Y=0.0313
MSU VQMT visualization
Original, Blurring (delta)-Y=0.0313
JPEG Q=80, Blurring (delta)-Y=0.0318
MSU VQMT visualization
Choose example:
Original, Blurring (delta)-Y=0.0175
LQ H264, Blurring (delta)-Y=0.0165
MSU VQMT visualization
Original, Blurring (delta)-Y=0.0175
Blurring, Blurring (delta)-Y=0.0107
MSU VQMT visualization
Original, Blurring (delta)-Y=0.0175
Random points, Blurring (delta)-Y=0.0186
MSU VQMT visualization
Original, Blurring (delta)-Y=0.0175
Luminance shift, Blurring (delta)-Y=0.0175
MSU VQMT visualization
Original, Blurring (delta)-Y=0.0175
JPEG Q=2, Blurring (delta)-Y=0.0169
MSU VQMT visualization
Original, Blurring (delta)-Y=0.0175
JPEG Q=5, Blurring (delta)-Y=0.0181
MSU VQMT visualization
Original, Blurring (delta)-Y=0.0175
JPEG Q=10, Blurring (delta)-Y=0.0185
MSU VQMT visualization
Original, Blurring (delta)-Y=0.0175
JPEG Q=15, Blurring (delta)-Y=0.0184
MSU VQMT visualization
Original, Blurring (delta)-Y=0.0175
JPEG Q=20, Blurring (delta)-Y=0.0184
MSU VQMT visualization
Original, Blurring (delta)-Y=0.0175
JPEG Q=40, Blurring (delta)-Y=0.0182
MSU VQMT visualization
Original, Blurring (delta)-Y=0.0175
JPEG Q=80, Blurring (delta)-Y=0.0181
MSU VQMT visualization

Legacy notes

In VQMT 12 metric was optimized, consider correct range, added legacy mode

MSU Blocking Metric

General info

Metric type:
no-reference image metric
Value range:
(no blocks) 0..∞ (a lot of blocks)
Value interpretation:
bigger is more blocks
MSU VQMT implementations:
CPU multithreaded
MSU VQMT visualization:
pixel-wise
Available colorspaces:
Y
Output values:
metric value
Aggregated values:
standard set
MSU VQMT usages:
-metr blocking

Algorithm description

This metric contains heuristic method for detecting objects edges, which are placed to the edge of the block. In this case metric value is pulled down, allowing to measure blocking more precisely. This metric also considers image contrast around of block and use it as weight for obtained value.

Notes:

  • This algirithm considers 8x8 blocks of image, so it applicable only for I frames of video and some of encoders.

Benchmark

Implementation System & settings Resulution FPS Sec. per frame
VQMT 13 default-Y 8 cores HD 720p 403.223 0.002
VQMT 13 default-Y 1 core enabled HD 720p 117.668 0.008
VQMT 13 default-Y 8 cores FullHD 1080p 182.505 0.005
VQMT 13 default-Y 1 core enabled FullHD 1080p 56.020 0.018
VQMT 13 default-Y 8 cores UHD 4k 2160p 44.448 0.022
VQMT 13 default-Y 1 core enabled UHD 4k 2160p 14.029 0.071
measurements were done on VQMT 12 PRO for Windows. Can vary depending on system configuration, input format and other factors

Example

Choose example:
Original, Blocking-Y=12.83
LQ H264, Blocking-Y=15.65
MSU VQMT visualization
Original, Blocking-Y=12.83
Blurring, Blocking-Y=16.77
MSU VQMT visualization
Original, Blocking-Y=12.83
Random points, Blocking-Y=18.58
MSU VQMT visualization
Original, Blocking-Y=12.83
Luminance shift, Blocking-Y=12.83
MSU VQMT visualization
Original, Blocking-Y=12.83
JPEG Q=2, Blocking-Y=243.89
MSU VQMT visualization
Original, Blocking-Y=12.83
JPEG Q=5, Blocking-Y=185.01
MSU VQMT visualization
Original, Blocking-Y=12.83
JPEG Q=10, Blocking-Y=94.45
MSU VQMT visualization
Original, Blocking-Y=12.83
JPEG Q=15, Blocking-Y=56.61
MSU VQMT visualization
Original, Blocking-Y=12.83
JPEG Q=20, Blocking-Y=41.44
MSU VQMT visualization
Original, Blocking-Y=12.83
JPEG Q=40, Blocking-Y=23.99
MSU VQMT visualization
Original, Blocking-Y=12.83
JPEG Q=80, Blocking-Y=14.60
MSU VQMT visualization
Choose example:
Original, Blocking-Y=10.08
LQ H264, Blocking-Y=16.16
MSU VQMT visualization
Original, Blocking-Y=10.08
Blurring, Blocking-Y=16.98
MSU VQMT visualization
Original, Blocking-Y=10.08
Random points, Blocking-Y=12.28
MSU VQMT visualization
Original, Blocking-Y=10.08
Luminance shift, Blocking-Y=10.06
MSU VQMT visualization
Original, Blocking-Y=10.08
JPEG Q=2, Blocking-Y=218.31
MSU VQMT visualization
Original, Blocking-Y=10.08
JPEG Q=5, Blocking-Y=153.01
MSU VQMT visualization
Original, Blocking-Y=10.08
JPEG Q=10, Blocking-Y=75.26
MSU VQMT visualization
Original, Blocking-Y=10.08
JPEG Q=15, Blocking-Y=44.29
MSU VQMT visualization
Original, Blocking-Y=10.08
JPEG Q=20, Blocking-Y=30.92
MSU VQMT visualization
Original, Blocking-Y=10.08
JPEG Q=40, Blocking-Y=16.21
MSU VQMT visualization
Original, Blocking-Y=10.08
JPEG Q=80, Blocking-Y=9.45
MSU VQMT visualization
Choose example:
Original, Blocking-Y=24.51
LQ H264, Blocking-Y=16.65
MSU VQMT visualization
Original, Blocking-Y=24.51
Blurring, Blocking-Y=16.47
MSU VQMT visualization
Original, Blocking-Y=24.51
Random points, Blocking-Y=33.65
MSU VQMT visualization
Original, Blocking-Y=24.51
Luminance shift, Blocking-Y=24.50
MSU VQMT visualization
Original, Blocking-Y=24.51
JPEG Q=2, Blocking-Y=173.27
MSU VQMT visualization
Original, Blocking-Y=24.51
JPEG Q=5, Blocking-Y=140.61
MSU VQMT visualization
Original, Blocking-Y=24.51
JPEG Q=10, Blocking-Y=71.47
MSU VQMT visualization
Original, Blocking-Y=24.51
JPEG Q=15, Blocking-Y=50.02
MSU VQMT visualization
Original, Blocking-Y=24.51
JPEG Q=20, Blocking-Y=38.73
MSU VQMT visualization
Original, Blocking-Y=24.51
JPEG Q=40, Blocking-Y=25.58
MSU VQMT visualization
Original, Blocking-Y=24.51
JPEG Q=80, Blocking-Y=19.53
MSU VQMT visualization

Legacy notes

In VQMT 12 metric was optimized, consider correct range, added legacy mode.

Spatial Information (SI)

General info

Metric type:
no-reference image metric
Value range:
(simple, monotone frame) 0..1 (very complex frame)
Value interpretation:
bigger more complex frame
MSU VQMT implementations:
CPU multithreaded
MSU VQMT visualization:
pixel-wise, sobel transformation
Available colorspaces:
Y
Output values:
metric value
Aggregated values:
standard set
MSU VQMT usages:
-metr si
External links:
ITU-T Recommendation P.910: Subjective video quality assessment methods for multimedia applications, 1999. - 37 p.

Algorithm description

This metric measures complexity (entropy) of an input image. This metric represents simplest SI realization that takes standard deviation of sequence of pixel values (Y-component) of Sobel transformation of input image:

is length of vector , where and are horizontal and vertical Sobel transformation. While calculating Sobel, the edge pixels will be excluded from calculation.

Benchmark

Implementation System & settings Resulution FPS Sec. per frame
VQMT 13 default-Y 8 cores HD 720p 339.146 0.003
VQMT 13 default-Y 1 core enabled HD 720p 95.112 0.011
VQMT 13 default-Y 8 cores FullHD 1080p 149.358 0.007
VQMT 13 default-Y 1 core enabled FullHD 1080p 44.072 0.023
VQMT 13 default-Y 8 cores UHD 4k 2160p 37.605 0.027
VQMT 13 default-Y 1 core enabled UHD 4k 2160p 11.157 0.090
measurements were done on VQMT 12 PRO for Windows. Can vary depending on system configuration, input format and other factors

Example

Choose example:
Original, SI-Y=0.0335
LQ H264, SI-Y=0.0233
MSU VQMT visualization
Original, SI-Y=0.0335
Blurring, SI-Y=0.0102
MSU VQMT visualization
Original, SI-Y=0.0335
Random points, SI-Y=0.0364
MSU VQMT visualization
Original, SI-Y=0.0335
Luminance shift, SI-Y=0.0335
MSU VQMT visualization
Original, SI-Y=0.0335
JPEG Q=2, SI-Y=0.0380
MSU VQMT visualization
Original, SI-Y=0.0335
JPEG Q=5, SI-Y=0.0348
MSU VQMT visualization
Original, SI-Y=0.0335
JPEG Q=10, SI-Y=0.0338
MSU VQMT visualization
Original, SI-Y=0.0335
JPEG Q=15, SI-Y=0.0333
MSU VQMT visualization
Original, SI-Y=0.0335
JPEG Q=20, SI-Y=0.0333
MSU VQMT visualization
Original, SI-Y=0.0335
JPEG Q=40, SI-Y=0.0335
MSU VQMT visualization
Original, SI-Y=0.0335
JPEG Q=80, SI-Y=0.0336
MSU VQMT visualization
Choose example:
Original, SI-Y=0.0631
LQ H264, SI-Y=0.0244
MSU VQMT visualization
Original, SI-Y=0.0631
Blurring, SI-Y=0.0203
MSU VQMT visualization
Original, SI-Y=0.0631
Random points, SI-Y=0.0650
MSU VQMT visualization
Original, SI-Y=0.0631
Luminance shift, SI-Y=0.0631
MSU VQMT visualization
Original, SI-Y=0.0631
JPEG Q=2, SI-Y=0.0634
MSU VQMT visualization
Original, SI-Y=0.0631
JPEG Q=5, SI-Y=0.0610
MSU VQMT visualization
Original, SI-Y=0.0631
JPEG Q=10, SI-Y=0.0602
MSU VQMT visualization
Original, SI-Y=0.0631
JPEG Q=15, SI-Y=0.0603
MSU VQMT visualization
Original, SI-Y=0.0631
JPEG Q=20, SI-Y=0.0603
MSU VQMT visualization
Original, SI-Y=0.0631
JPEG Q=40, SI-Y=0.0614
MSU VQMT visualization
Original, SI-Y=0.0631
JPEG Q=80, SI-Y=0.0624
MSU VQMT visualization
Choose example:
Original, SI-Y=0.0625
LQ H264, SI-Y=0.0591
MSU VQMT visualization
Original, SI-Y=0.0625
Blurring, SI-Y=0.0264
MSU VQMT visualization
Original, SI-Y=0.0625
Random points, SI-Y=0.0649
MSU VQMT visualization
Original, SI-Y=0.0625
Luminance shift, SI-Y=0.0625
MSU VQMT visualization
Original, SI-Y=0.0625
JPEG Q=2, SI-Y=0.0717
MSU VQMT visualization
Original, SI-Y=0.0625
JPEG Q=5, SI-Y=0.0652
MSU VQMT visualization
Original, SI-Y=0.0625
JPEG Q=10, SI-Y=0.0630
MSU VQMT visualization
Original, SI-Y=0.0625
JPEG Q=15, SI-Y=0.0626
MSU VQMT visualization
Original, SI-Y=0.0625
JPEG Q=20, SI-Y=0.0625
MSU VQMT visualization
Original, SI-Y=0.0625
JPEG Q=40, SI-Y=0.0624
MSU VQMT visualization
Original, SI-Y=0.0625
JPEG Q=80, SI-Y=0.0624
MSU VQMT visualization

Legacy notes

This metric firstly imlemented in VQMT 11, then excluded in VQMT 12 and returned in VQMT 13

Spatial Information (TI)

General info

Metric type:
no-reference temporal metric
Value range:
(simple, static video) 0..1 (very diverse frames)
Value interpretation:
bigger more diverse frames
MSU VQMT implementations:
CPU multithreaded
MSU VQMT visualization:
pixel-wise, difference between adjacent frames
Available colorspaces:
Y
Output values:
metric value
Aggregated values:
standard set
MSU VQMT usages:
-metr ti
External links:
ITU-T Recommendation P.910: Subjective video quality assessment methods for multimedia applications, 1999. - 37 p.

Algorithm description

This metric measures complexity (entropy) of defference between consequent frames of input video. This metric represents simplest TI realization that takes standard deviation of sequence of differences of corresponding pixel values of a frame and previous frame:

Benchmark

Implementation System & settings Resulution FPS Sec. per frame
VQMT 13 default-Y 8 cores HD 720p 691.497 0.001
VQMT 13 default-Y 1 core enabled HD 720p 252.300 0.004
VQMT 13 default-Y 8 cores FullHD 1080p 271.375 0.004
VQMT 13 default-Y 1 core enabled FullHD 1080p 98.760 0.010
VQMT 13 default-Y 8 cores UHD 4k 2160p 67.874 0.015
VQMT 13 default-Y 1 core enabled UHD 4k 2160p 25.194 0.040
measurements were done on VQMT 12 PRO for Windows. Can vary depending on system configuration, input format and other factors

Example

Choose example:
Original, TI-Y=0.0147
LQ H264, TI-Y=0.0033
MSU VQMT visualization
Original, TI-Y=0.0147
Blurring, TI-Y=0.0037
MSU VQMT visualization
Original, TI-Y=0.0147
Random points, TI-Y=0.0233
MSU VQMT visualization
Original, TI-Y=0.0147
Luminance shift, TI-Y=0.0147
MSU VQMT visualization
Original, TI-Y=0.0147
JPEG Q=2, TI-Y=0.0192
MSU VQMT visualization
Original, TI-Y=0.0147
JPEG Q=5, TI-Y=0.0199
MSU VQMT visualization
Original, TI-Y=0.0147
JPEG Q=10, TI-Y=0.0186
MSU VQMT visualization
Original, TI-Y=0.0147
JPEG Q=15, TI-Y=0.0180
MSU VQMT visualization
Original, TI-Y=0.0147
JPEG Q=20, TI-Y=0.0171
MSU VQMT visualization
Original, TI-Y=0.0147
JPEG Q=40, TI-Y=0.0160
MSU VQMT visualization
Original, TI-Y=0.0147
JPEG Q=80, TI-Y=0.0152
MSU VQMT visualization
Choose example:
Original, TI-Y=0.0004
LQ H264, TI-Y=0.0127
MSU VQMT visualization
Original, TI-Y=0.0004
Blurring, TI-Y=0.0005
MSU VQMT visualization
Original, TI-Y=0.0004
Random points, TI-Y=0.0206
MSU VQMT visualization
Original, TI-Y=0.0004
Luminance shift, TI-Y=0.0004
MSU VQMT visualization
Original, TI-Y=0.0004
JPEG Q=2, TI-Y=0.0021
MSU VQMT visualization
Original, TI-Y=0.0004
JPEG Q=5, TI-Y=0.0016
MSU VQMT visualization
Original, TI-Y=0.0004
JPEG Q=10, TI-Y=0.0011
MSU VQMT visualization
Original, TI-Y=0.0004
JPEG Q=15, TI-Y=0.0011
MSU VQMT visualization