Fractal Domain-based Color Image Retrieval Research

With the development of information technology and popularization in image compression, the content-based image retrieval in compress domain becomes the focus of research. The image is denoted with Iterative Function System of fractal method, so it has been widely used in image compression, as well as it could be used in the content-based image retrievalTo solve the problem of the content-based color image retrieval in fractal domain, a matching algorithm is proposed in this paper, which is based on the score matrix and luminance parameters in fractal code. Firstly each color image in the database is compressed by fast fractal coding and IFS fractal code is got, then the index features including self-similarity feature and color feature are extracted from the fractal code for image retrieval. The work has been done in this paper is listed below.(1) At the image fractal coding phase, a deferring range/domain comparison fractal compression coding method for color image is proposed. Using the correlation of r, g, b components of a color image, it merges three color components into one. The strategy of postponing range\domain comparisons with respect to a preset block is utilized to search a close match of the range block, then the luminance coefficients are calculated again, the IFS fractal code of color image is finally achieved.(2) At the image retrieval phase, an algorithm of image retrieval based on the score matrix is presented. It is completely completed upon the fractal compression domain. The score matrix is established in term of the position of the matched domain block for each range block, and statistical numbers are acquired from the partition of matrix in polar coordinates, then the image retrieval is performed.(3) At the image retrieval phase, combining the acquired fractal code and the algorithm based on the score matrix, an algorithm of color image retrieval is proposed further based on the score matrix and luminance parameters. Firstly, the parameters are extracted from the code. Secondly, they are combined with the features obtained from score matrix by specifically weight value. Finally the index features have been got and the color image retrieval is executed.(4) At the experiment phase, the method of compression proposed in this paper is analyzed and compared with other methods in three aspects including compression ratio, compression rate and quality of decoded image, in the same way, the two algorithms of retrieval in this paper are compared with the others in two aspects including complexity and retrieval ratio. Because of adding the color feature to the index features, the acquired retrieval ratio is improved, in other words, the second method is superior to the first one in this paper.Experiments shows that the fractal compression method for color image presented in this paper improves the fractal coding speed further, and the algorithm of color image retrieval based on the score matrix and luminance parameters can increase the retrieval ratio efficiently.