Image Retrieval Algorithm Based On Fractal

The extraction of fractal code and comparison between similarities are the key technologies. To achieve real time and accuracy retrieval, this paper discussed different fractal algorithms in image retrieval. The test result of histogram retrieval method is given as the standard of accurate-complete retrieval rating in future works; the test result of basic fractal encoding is given as the standard of PSNR in future works. Two partial code retrieval methods, contiguous-match and HV segmentation, are put forward.The segmentation matching strategy in quadtree fractal algorithm are restricted in this paper, the D blocks are selected in four directions only from the neighbors, to avoid mass searching. The image encoding speed is improved. The comparison algorithm of encoded IFS is analyzed, the size, location and matching direction between different R blocks are considered, and a new distance formula is put forward.The HV fragmentation based fractal retrial algorithm improves the image adaptability of fragmentations. Only one encode matching is required for each R block, so the encoding efficiency is further improved. For different images, the size and ratio of blocks are different, a weight formula based on cross area is put forward to compare the similarities of different R blocks, a new retrial formula based on HV fragmentation algorithm is obtained.According to the algorithm given above, this paper test the encoding and decoding of 256×256 size grey images. Under certain testing conditions, the encoding time, decoding quality and fragmentation effect are especially analyzed. Both Quadtree and HV segmentation fractal retrieval algorithm can improve the encoding speed over 40 times, and the decoding quality is guaranteed. In the test of complicated texture images, the HV based fractal retrieval algorithm has higher decoding quality over the original algorithm. To test the retrieval effect against different image types, the paper test the character, scenery, texture and daily-life images. The results show that the quadtree and HV based fractal retrieval algorithm can embody the similarity of image structure, and which is prior than the histogram retrieval algorithm.