Perspective distortions estimation


Sometimes, during stereoscopic movie shooting cameras can be installed incorrectly, resulting in a variety of geometric distortions. Particular case of geometric distortions is an artifact that occurs when stereoscopic cameras are vertically misaligned. We propose a new algorithm to detect a mismatch in the vertical position of the cameras.


Proposed method

Due to the geometric properties of optics, objects decrease in size proportionally to the distance between the object and camera. This phenomenon is called a linear perspective. In this regard, the algorithm is based on the assumption that, if the vertical position of the cameras does not match, the vertical shift value changes proportionally to the distance.

The algorithm is based on motion estimation and a modified version of RANSAC.

Algorithm scheme


For the analysis of stereoscopic frames where the artifact was not expressed explicitly, compensation was applied — the horizontal shift of the right image so that the background objects had horizontal disparity values equal to zero.

Left image


Compensated difference

In this work, 14 stereoscopic movies were analyzed and 122 scenes with this artifact were found.

The efficiency of this method was proven on the artificially created test set, the average confidence in the results was approximately 90%. Pearson correlation coefficient between ground truth and algorithm values was 0.993.

Experimental evaluation

A validation dataset was also created, consisting of 200 scenes taken from 4 stereo films.



The average speed of the algorithm on the Intel Core i7-6500U 2.50GHz CPU is 1.04 seconds per 960 × 544 resolution frame with 96.8% of the time spent on calculating the disparity and confidence maps.

05 May 2020
See Also
Call for HEVC codecs 2019
Fourteen modern video codec comparison
Parallax range estimation in S3D video
Parallax determines the depth of S3D movies. The range of parallaxes should be both comfortable and entertaining for spectators.
Geometric distortions analysis and correction
Production of low-budget movies is prone to errors. Our method automatically corrects rotation and scale mismatch.
Automatic detection of artifacts in converted S3D videos
Our set of algorithms detects edge sharpness mismatch, cardboard effect, and crosstalk noticeability.
Temporal shift estimation for stereoscopic videos
How to take into account geometric distortions in the estimation of the temporal shift?
Neural network-based algorithm for classification of stereoscopic video by the production method
What method was used to create the 3D scene?
Site structure