Image stitch-based multi-view video coding

Abstract

In general, it is necessary for Multi-view Video Coding (MVC) methods to compress multi-view videos efficiently and have a property of view-scalability in order to decode arbitrary views according to any viewer's interests. Much research has been done on MVC methods, with the goal of increasing coding efficiency. Although these previous methods have considered the property of view-scalability, a lot of coding bits and delays were necessary to decode arbitrary views. In this paper, we propose an MVC method based on image stitching. We generated a stitched reference and encoded multi-view sequences using disparity-compensated method. The proposed method is able to reduce delays during the decoding stage. Experimental results show that the proposed MVC method increased the PSNR by 1.5 similar to 2.0dB and saved 10% of the coding bits compared to simulcast coding.