Joint Multiview Video Plus Depth Coding with Saliency Approach

Abstract

The main consideration of compression efficiency in multiview is to venture the temporal and interview analytical dependencies since all cameras capture the particular frame with variable viewpoints. The era of multiview coding (MVC) predictor selection statistics is used to compress MVD representation resulting in separate bitstreams of texture and video sequences. These coding schemes do not capture the similarities that arise in texture and depth video sequence, whereas joint multiview video plus depth coding (JMVDC) scheme employs the correlation co-efficiency of the motion in texture and depth sequence in rendering the object of interest in the scene. Large amount of data that is produced in this representation becomes a challenge for data storage and network transmission. In JMVDC, the structure enables interlayer motion prediction mechanism by representing the base and enhancement layers as texture and depth. The inter-dependency of motion texture and depth content of multiview is accomplished and achieved by employing different correlation coefficient methods. The proposed method utilizes depth map to enhance the salient region with the combination of local and global saliency information. The experimental results show that the JMVDC method in saliency map with enhancement achieves the reduction of computational complexity to detect the distinctive region.