Adaptive compressed sensing for depthmap compression using graph-based transform

Sungwon Lee, Antonio Ortega
2012 2012 19th IEEE International Conference on Image Processing  
Depthmap has sharp edges where depth changes. Precise edge information crucial in depth image based rendering (DIBR) system Figure: Multiview video puls depth (MVD) using depth image based rendering(DIBR) [1] Figure: Multiview video plus depth example [2] Error in edge information → geometry error in interpolated view → significant degradation of the quality [3] Traditional DCT (H.264/AVC) not efficient for arbitrary edges → requires edge-adaptive transform Related Work Graph-based transform
more » ... T) [3] Represent depth map as a graph Present edge-adaptive GBT construction Propose a greedy algorithm to find an optimal GBT CS-based depthmap compression [4, 5] Use fixed DCT as sparsifying basis Not adaptive to edges in blocks; inefficient for arbitrary edges GBT for Depthmap Compression GBT efficiently sparsify arbitrary edges in depthmap [6] Construction procedure For a given graph, G(V, E), construct adjacency matrix, A Compute Laplacian matrix, L:
doi:10.1109/icip.2012.6467013 dblp:conf/icip/LeeO12 fatcat:cq2j226olbc3tolrwd4emiay5e