Coding of arbitrarily shaped image segments based on a generalized orthogonal transform

摘要: Abstract Region oriented image representation offers several advantages over block-oriented schemes, e.g. adaptation to the local characteristics, or object motion compensation as opposed block-wise compensation. For task of data compression, i.e. coding, new algorithms are needed which work on arbitrarily shaped regions, called segments, instead rectangular blocks. Based a generalized moment approach, luminance function inside segment is approximated by weighted sum basis functions, for example polynomials. A set functions orthogonal with respect shape be coded can obtained using orthogonalization schemes. This results in derivation shape-adapted transform coder. Suitable coder and decoder structures introduced do not necessitate transmission each segment. Finally an application derived sequence coding at low rates shown, based segmentation compensated prediction error image.