Estimating Depth-Salient Edges And its Application To Stereoscopic Image Quality Assessment

Sameeulla Khan Md,Sumohana Channappayya

DOI: https://doi.org/10.48550/arXiv.1802.03883

2018-09-21

Abstract:The human visual system pays attention to salient regions while perceiving an image. When viewing a stereoscopic 3D (S3D) image, we hypothesize that while most of the contribution to saliency is provided by the 2D image, a small but significant contribution is provided by the depth component. Further, we claim that only a subset of image edges contribute to depth perception while viewing an S3D image. In this paper, we propose a systematic approach for depth saliency estimation, called Salient Edges with respect to Depth perception (SED) which localizes the depth-salient edges in an S3D image. We demonstrate the utility of SED in full reference stereoscopic image quality assessment (FRSIQA). We consider gradient magnitude and inter-gradient maps for predicting structural similarity. A coarse quality estimate is derived first by comparing the 2D saliency and gradient maps of reference and test stereo pairs. We refine this quality using SED maps for evaluating depth quality. Finally, we combine this luminance and depth quality to obtain an overall stereo image quality. We perform a comprehensive evaluation of our metric on seven publicly available S3D IQA databases. The proposed metric shows competitive performance on all seven databases with state-of-the-art performance on three of them.

Image and Video Processing