A Novel Fractal Block Coding Method by Using New Shape-based Descriptor

The basic theory of Jacquin’s fractal block coding is to approximate a natural image by its partial subdivided parts, given that the parts resemble the whole at either the same scale or different scales. The fact that self-similarity most commonly exists within an entire natural image leads to the development of the proposed method that uses arc as the fractal descriptor, in which an arc commonly exists within the entire natural image, named as Arc-Descriptor Fractal Coding (ADFC) method. In the ADFC, each range block is approximated by using the selected arc-descriptor from an optimal pool. This paper experimentally demonstrates the ADFC method on Java and the ADFC system is verified with 5 medical images. The experimental results indicate the ADFC method can encode and decode the experimental images effectively. The PSNR of the images after encoding can reach 30 dB at limited cost while the CR is less than 17%. The ADFC outperforms Jacquin’s method under the comparison of the number of search in matching process. We conclude that the ADFC successfully encodes and decodes image in an efficient search during encoding phase without noticeable loss of image quality