A Generalized Method for Image Coding Using Fractal-Based Techniques

A new general method is proposed for image coding which exploits similarities, possibly with scaling, among different parts of the image. The coding is performed by approximating each image block with a linear combination of blocks selected from a library of blocks. This library is made up of (1) a set of fixed library blocks, (2) a set of blocks taken from the filtered, subsampled image, and (3) a set of blocks taken from the image without any change of scale. When the last two sets are selected causally, the decoding process is noniterative with no constraints on the coefficients of the library blocks. The index of the selected library blocks and their corresponding coefficients make the code for each range block. Methods are proposed for making the library and selecting blocks from the library.