RUSSIAN

MSU Stereo To Multiview 3D Video Conversion - content creation technology for glasses-free 3D displays

Take a look at this article on the new site! Follow the link
https://videoprocessing.ai/video_filters/stereo-to-multiview.html

Project: Dr. Dmitriy Vatolin
Implementation, algorithm: Sergey Grishin,
Karen Simonyan,
Sergey Matyunin


Introduction


Now 3D visualization is a fast developing technology. Autostereoscopic glasses-free 3D monitors are already available. But video materials for them are still rare. Most of this video is rendered. 3D video creation is expensive and complicated task, because glasses-free 3D displays require several views of the same scene (usually 8-9 views). Stereo video capture is much easier. Stereo films become more and more popular. Our lab have developed stereo to multiview video conversion algorithm for glasses-free autostereoscopic 3D displays. This algorithm makes 3D video creation problem less complicated.

Stereo to multiview conversion

Work description


It is necessary to generate intermediate views to convert stereo to multiview intermediate.

We have developed conversion algorithm and processed several professionally captured stereo videos. We also researched 3D video capturing (more about MSU 3D Video Capture).

Intermediate views interpolation
Intermediate views interpolation

Intermediate views generation doesn't use depth map construction and semitransparent objects processing is correct. We also work on depth and disparity map generation. It can be used in monoscopic to 3D video conversion.

Our stereo video converting doesn't need iterative processes and it is suite for stream processing. Real-time realization is possible. We also work on GPU version of conversion for more acceleration.

It is also possible to create multiview or stereo video from monoscopic sequences. We developed a conversion algorithm that uses only temporal and spatial frame information to generate 3D video: MSU Depth Map Generation and Stereo video creation.


How to view 3D video


To view 3D images you need autostereoscopic monitor and specialized player. Depending on how many views does your monitor support you should download 5-view or 9-view video. View order is also important. You should choose view order (from left to right OR from right to left) supported by your player.

5 view alignment: 9 view alignment:
5 view alignment. From left to right. 9 view alignment. From left to right. 9 view alignment. From right to left.
From left to right From left to right From right to left
9-view frame
9-view frame

Conversion output format can be also specified to be a sequence of numbered views or other appropriate way.


Download


Here you can download our results.

MSU 3D video

MSU 3D Demo Video 5 Views. From left to right.
Frame 600

MSU 3D Demo Video 5 Views. From right to left.
Frame 600

MSU 3D Demo Video 9 Views. From left to right.
Frame 600

MSU 3D Demo Video 9 Views. From right to left.
Frame 600

All videos are encoded by XVID codec.

More about 3D video capturing: MSU 3D Video Capture

Other multiview video

5 view 9 view
From left to right From left to right From right to left

Skydiving1 ( 5 Views, left )
Frame 0

Skydiving1 ( 9 Views, left )
Frame 0

Skydiving1 ( 9 Views, right )
Frame 0

Skydiving2 ( 5 Views, left )
Frame 0

Skydiving2 ( 9 Views, left )
Frame 0

Skydiving2 ( 9 Views, right )
Frame 0

Skull_Rock ( 5 Views, left )
Frame 420

Skull_Rock ( 9 Views, left )
Frame 420

Skull_Rock ( 9 Views, right )
Frame 420

Heidelberg ( 5 Views, left )
Frame 100

Heidelberg ( 9 Views, left )
Frame 100

Heidelberg ( 9 Views, right )
Frame 100

Mouldpenny ( 5 Views, left )
Frame 0

Mouldpenny ( 9 Views, left )
Frame 0

Mouldpenny ( 9 Views, right )
Frame 0

Authors of films:
Skydiving - Eric Deren, Dzignlight Studios
Skull Rock - Red Star Studio
Heidelberg, Mouldpenny - Meinolf Amekudzi, Dongleware Verlags GmbH


Contacts


For questions and proposition please contact us  

Video resources:

Last updated: 12-May-2022


Server size: 8069 files, 1215Mb (Server statistics)

Project updated by
Server Team and MSU Video Group

Project sponsored by YUVsoft Corp.

Project supported by MSU Graphics & Media Lab