A novel image quality index for stereo image

Abstract

Stereo image quality assessment is provided as computa-tional methods to automatically assess the quality of stereo im-ages in perceptually consistent manner. In this paper, a novel full reference stereo image quality index with highlighting double-channel of binocular combination is introduced to assess stereo image quality. First, considering the visual sensitivity of stimu-lus at different spatial frequencies is different, both reference and distortion stereo pairs are filtered by a CSF (contrast sensitivity function). Second, in order to mimic the binocular fusion me-chanism of HVS (human visual system), we extract the binocular energy response of both reference and distortion stereo pairs based on a magnitude response of Gabor filtering measure. Third, since the HVS is enormously complex, especially the binocular rivalry which caused by the asymmetric distortions of stereo pair, we propose to fix the issue of dominant view by using a block-based contrast measure. Finally, a single estimate of overall per-ceived quality of stereo images is obtained by the qualities pool-ing of double-channel. Experimental results show that the pro-posed metric achieves significantly higher consistency with sub-jective scores.