A new fractal image coding scheme employing blocks of variable shapes

In the fractal coding schemes proposed so far based on iterated function systems, an image is represented by a self-affine set of square blocks. We propose a scheme of fractal image coding with blocks of variable shapes. In this scheme, the range blocks are determined by the so-called splitting-and-merging method. Also, we define two kinds of range blocks, shade blocks and non-shade blocks, according to the variance of the brightness level inside blocks. By this method, range blocks both of various size and of various shapes are extracted from an input image. The larger range blocks result in the a greater reduction of the total number of blocks and, hence, the amount of encoded bits. Because of the specified order used in merging of primitive blocks, only a few more bits are added for each range block to encode its shape. The results of computer simulation with a test gray-scale image show that compared with the conventional scheme using square range blocks, the proposed coding scheme reduces about 0.2 bits/pel to achieve almost the same reproduced quality