The Research On Image Compression Technology Based On Wavelete And Fractal

The principle of fractal image compression is that local self-similarity form a Local Iterated Function System. The image quality, compression ratio and code speed of fractal decode depend on the size of child block. If the size of child block is bigger, then the compression ratio and code speed is higher, but the image quality is poorer after decoding. On the contrary, the quality of the image is better after decoding, but the code speed is slower. Quadtrees algorithm can make the child block bigger in the region where it has good self-similarity. Otherwise, quadtrees algorithm makes the child block smaller . Thus the quadtrees algorithm can ensure good image quantity of fractal compression after decoding. The evolution algorithm is a random search algorithm. It is faster than method of exhaustion when we search the most matched parent block of a child block in fractal image compression. The two-dimentional grey level image is divided into the low-frequency area and high-frequency area after the wavelet transform ,and low-frequency area concentrates the most of energy of the image in the small area,and high-frequency area spreads little energy of the image in the wide area. The primitive local self-similarity in two-dimentional grey level image is destroied after wavelet transform. But wavelet coefficients exists self-similarity among the layeres in high-frequency area. So the image can be compressed by fractal after wavelet transform. This paper proposes a new still image compression algorithm based on fractal and wavelet. It uses different algorithms to compress wavelet coefficients because of the differences of the energy distributing between low-frequency area and high-frequency area. The wavelet coefficients in low-frequency area are saved directly after quantized; the self-similarity among layers form a Local Iterated Function System, so wavelet coefficients in high-frequency area can be compressed by fractal. The new algorithm uses quadtrees to divide the high-frequency area more reasonablely, and raise quality of the image after decoding; The new algorithm uses evolve algorithm to raise the match speed between father blocks and child block; The new algorithm use the validity scan of child block to let fractal code only in the valid region which the MSE is bigger than 2,the validity scan make the child block become smaller but it don't increase the number of child block, it raise the efficiency of fractal compression in wavelet. The result indicates this algorithm is successful.