An efficient parameter selection scheme for view level rate-distortion control in multi-view/3D video coding

Abstract

Multi-view/3D video technology has made noticeable progress in recent years and has opened its way in entertainment, education, security, and surveillance applications. A multi-view/3D video expands user's experience beyond the conventional single view video by showing the same scene captured simultaneously from multiple cameras from different positions and through different angles. One of the major challenges in this field is how to transmit the large amount of data of a multi-view/3D sequence. Rate control is one of the key components of any multi-view/3D video encoder and has a significant impact on its performance in terms of Rate-Distortion (RD). Rate control usually exploits a RD model to describe the relationship between the rate and the quality of the encoded video. The precision and computational complexity of a RD model is affected by the RD parameters used in rate-distortion modeling. In this paper, an efficient parameter selection scheme for view-level RD models has been proposed that can improve the efficiency of view-level RD models in terms of prediction precision and computational complexity. Experimental results indicate that using the extracted RD parameters of our proposed approach, the rate of each view can be predicted with relatively high precision and a low estimation error of 12% on average.