Depth map compression based on platelet coding and quadratic curve fitting

In this paper, we propose an efficient depth data compression scheme for depth images. Our proposed scheme is a platelet based coding method using quadtree decomposition, Lagrangian optimization function, and quadratic curve fitting algorithm. In this scheme, quadtree decomposition is used to divide a depth map into blocks with one or two homogeneous regions; Lagrangian optimization is employed to minimize the distortion as well as to keep relative high compression ratio within each sub-block; Four quadratic curve fitting models are designed for the objects´ contours in order to keep the edges´ position more precisely after compression. The experimental results show that our proposed method can achieve compression ratio comparable to the other known platelet based method. Besides, the objects´ contours are kept better compared to the other method.