Successive Interpolation/extrapolation Scheme for Missing Segments Restoration

Lately, great concern in image processing is devoted to region-oriented methods. Region-oriented image representation offers several advantages over block-oriented approach, e.g. adaptation to the local image characteristics. New algorithms are necessary for image coding, if we work with arbitrarily shaped image regions, called segments, instead on rectangular blocks. The original approach for the coding of arbitrarily shaped image segments based on a generalized orthogonal transform was discussed in [1]. Application scheme with cosine transform is proposed in [2]. The coding method used in this paper was thoroughly described in [3]. The transmission of images coded by block or segment based techniques via a noisy channel may lead to block or segment loss. Therefore error detection and concealment at the decoder side has to be applied. Commonly, spatial error concealment is used. It utilizes the surrounding correctly received image information to restore the damaged or missing pixels. A standard approach [4] assumes that the image content is changing smoothly. Hence the algorithm tries to restore the transition across the block boundary as smooth as possible. The extrapolationbased method of [5] tries to reconstruct the missing pixels as a weighted linear combination of correctly received pixels. Hence, the method can lead to solving a great number of linear equations and is computationally very complex. The transmission of block-coded image data in via wireless channel is described in [6]. Very interesting and novel method for spatial error concealment based on successive extrapolation of missing blocks is described in [7].