Region-oriented transform coding (ROTC) of images
Michael Gilge · International Conference on Acoustics, Speech, and Signal Processing · 2002
The method of transform coding for image-data compression is generalized by regarding transform coding as a least-squares approximation of two-dimensional functions. By an orthogonalization of basis functions with respect to a particular segment shape, a generalized transform coding scheme is derived. The algorithm contains all block-oriented transforms as special cases and allows the construction of new transforms, e.g. polynomial or spline transforms. The theoretical results are converted into coder and decoder structures enabling region-oriented transform coding without a transmission of the orthogonal basis functions for each segment.>