Automatic sharpness mismatch detection and compensation in stereo

Algorithm works frame-wise: for a pair of source image’s views it outputs whether it contains significant sharpness mismatch or no, and tries to authomatically fix it. Applied to every frame of video, it automatically detects all frames in source video with such a problem and fixes them.

Algorithm consists of three steps:

Computational speed: 1/16 FullHD stereo frames per second on Intel Core i5 + 4Gb RAM

Picture below demonstrates performance of proposed algorithm over multiple video sequences. For each sequence, averaged metric from step two of the algorithm is calculated using source and resulting stereo video

Algorithm results

Algorithm results

Algorithm is compared with commercial plug-in for Nuke, Ocula 3.0 (FocusMatcher) over set of 8 stereo sequences.

Pros over Ocula 3.0:

Results comparision with Ocula 3.0 (FocusMatching)

Results comparision with Ocula 3.0 (FocusMatching)

Pictures below illustrate differences between proposed algorithm and FocusMatcher on a single frame. It is notable that FocusMarcher corrupts background behind semi-transparent object (red boxes), while proposed algorithm doesn’t have such effect

Example stereo frame

Example stereo frame

Results of proposed algorithm

Results of proposed algorithm

Ocula 3.0's results (red box - produced artifacts)

Ocula 3.0's results (red box - produced artifacts)
27 May 2019
See Also
Call for HEVC codecs 2019
Fourteen modern video codec comparison
Automatic local color correction in S3D video
Stereo video may contain a huge color discrepancy. Most of the problems are hard to eliminate because of possible different distortions in each area in the frame.
Swapped views detection in S3D movies
Channel mismatch is one of the most painful-for-viewer problems in stereo 3D movies. This problem is very hard to detect and simply to eliminate: just swap the views.
MSU 3D-video Quality Analysis (MSU Video Quality Measurement Tool 3D Project: VQMT3D)
MSU 3D-video Quality Analysis. Report 9
MSU 3D-video Quality Analysis. Report 8
Site structure