Asymmetric Coding of Stereoscopic 3D Images with Perceptual Quality Control

Abstract

Significant research works have been devoted to asymmetric stereoscopic coding technique. Based on the perceptual theory of binocular suppression, in this method, one view of the stereo pair can be coded with a lower quality than the other, thus allowing to decrease bandwidth requirement for 3D visual content delivery without degrading subjective quality. However, with the condition that the quality gap between both images does not exceed a certain threshold. Some works proposed to fix this threshold by subjective experiments. Nevertheless, using a fixed threshold does not allow adaptation to the stereo image content and presents dependency to the experimental design. In this paper, we propose a novel method that selects automatically the maximum tolerable quality gap between views for different stereo image pairs. This has been achieved by modeling the relationship between the asymmetric quality and the inter-view distortion between corresponding pixels. Then, the derived relationship is combined with Binocular Just Noticeable Difference (BJND) model to control the quality threshold between the left and right views. Experimental results showed that the proposed coding method can achieve considerable bitrate saving up to 16.32% and outperforms the widely used asymmetric coding approaches.