COMPRESSION RATIO IN FRACTAL ENCODING OF STILL IMAGES

Abstract

Fractal encoding (compression) of halftone images is based on the hypothesis that any image can reveal the local self-image of its various parts. In fractal compression, images are encoded and broken into many non-overlapping blocks (rank regions), for each of which, within the same image, a larger block (domain) is searched whose pixels would be translated by some coefficient given by several coefficients. in the pixel of the rank area. In this case, the code will be the location and size of the rank areas, as well as the conversion factors that describe self-similarity within the image. In this paper, the fractal methods of compressing still images are investigated for transmission in multimedia applications. A modified search algorithm for similar blocks in the image is proposed, which differs from the existing by parallel simultaneous comparison of rank blocks in all domain-based indicators. Domain and rank blocks are described by feature vector. The feature vector includes the mean deviation in the block, the asymmetry in the block, the difference in contrast between the pixels in the block, the average deviation of the pixel brightness within the domain (rank) block. The deviation error allows you to adjust the compression ratio. If the desired quality is not achieved, then continue to reduce the size of or rank, or domain blocks. The code sequence is formed from the brightness values and color difference signals of the rank block, its feature vector and block addresses that are similar to it. The dependences of the compression ratio and the signal-to-noise ratio on the standard deviation of the average brightness of the rank blocks of the descendants on the brightness of the original rank block are obtained.