Lossy-to-lossless progressive coding of depth-maps

A progressive coding method is proposed for depth-map images, where the bitstream is encoded so that one can generate many lossy versions of the original, encompassing a wide range, from very low resolution up to lossless reconstruction. The partitions into regions of the lossy versions are assumed to be nested, so that a higher resolution image is obtained by splitting some regions of a lower resolution image. The encoder transmits to the decoder information about which regions to split, the extra contour to be added for obtaining the shapes of the more refined regions, and the extra depth values needed inside each new region. The efficient encoding of the anchor points in the progressive scenario, relative to the contour points already encoded, and the depth information recovery, are the main contributions of this paper. The progressive bitstream produced by the proposed method scales well over the whole range of rates, from low rates to lossless, reaching a performance close to that of the non-progressive methods.