Region of interest oriented fast mode decision for depth map coding in DIBR

During the depth image based rendering (DIBR), depth map is a new type of video sequence to be compressed. The synthesized result of DIBR is highly related to the quality of the depth map. Depth map also has some new features than color maps. It occurs to us that some improvement is required. Higher efficiency and keeping the edge quality is two critical issues for depth compression. In this paper, region of interest (ROI) detection is introduced to meet these requirements. We adopt an ROI detection method which can distinguish the moving and relatively complexity part (foreground) from each frame. By using different mode decision strategy in ROI and non-ROI, computing burden is reduced and performance is kept. Experimental results show that the proposed strategy saves 40% computation of ME on average with an acceptable performance lost.