Automatic analysis of sharpness mismatch between stereoscopic views for stereo 3D videos

This paper presents an efficient approach to measure sharpness mismatches between stereoscopic views. Sharpness mismatch can occur through focus mismatches between stereoscopic cameras, errors in post-processing or even for low-bandwidth transmission, where one view is subsampled or transmitted at a much lower rate. This artifact can lead to a degraded 3D experience for observers. In this paper, the sharpness mismatch score is estimated by measuring the width deviations of edge pairs in each valid depth plane. The mismatch probability is then calculated considering the perceptibility of edge width deviations. In the experiments, Gaussian low-pass filters were used to generate global sharpness mismatches between stereoscopic views since defocus-based effects of lens aberrations can be modeled as Gaussian blur. Thus, the global sharpness distortions simulate the focus mismatch of stereo cameras. The disparity maps of test videos were automatically generated and corrected. In addition, original high-quality disparity maps of the test datasets were used as benchmarks. According to the experimental results, we show that the proposed approach performs well on measuring sharpness mismatch between stereoscopic views by comparison with some state-of-the-art metrics.