Compressing depth maps using multiscale recurrent pattern image coding

Abstract

The use of the multidimensional multiscale parser algorithm for depth maps coding is proposed. The compression method uses a block-based approach, where efficient prediction combined with pattern matching is applied to the encoding of greyscale images, which convey the disparity or depth information of a 3D image. Simulation results show gains of up to 10 dB when compared with state-of-the-art methods, such as JPEG2000 and H.264/AVC.