Error-robustness of polar contour-coding

Abstract

For communication using digital video, compression is mandatory because of the high bit-rate requiring a large bandwidth or storage capacity. Recent developments as well as upcoming standards like MPEG-4, use not images but so-called video objects having an arbitrary shape. The contour of this shape has to be transmitted and therefore compressed as well. In mobile video communication transmission errors and loss of data will occur and this means that the compression has to be error-robust. This paper discusses the errorrobustness of a technique for contour coding using polar coordinates and the Discrete Cosine Transform (DCT). The recently proposed [11] polar technique transforms the Cartesian x and y contour coordinate functions to the domain of polar coordinates yielding an r and a φ function. Subsequently these functions are transformed using a DCT and quantised. We investigate the impact of not-received DCT coefficients on the reconstructed intra- and intercoded contours. When inter-coding, the contour with respect to which the current contour is coded, is composed of a weighed sum of N previous contours to reduce the impact of errors. Results show that for intercoded contours, this technique can be considered error-robust, except for errors in the header. For highly compressed intra-coded contours this is only partly so due to the high energy in most of the transmitted coefficients, which have to be well protected like the header of the contour information.