The Realization Of The Image Retrieval System Based On Fractal Technology

With the rapid growth of multimedia information data, more and more image information has blend in people’s lives. How to store and analyze the image information effectively has become a hot research topic. In recent years, the rapid development of fractal technology and proposed a relatively mature field compression algorithms compress, also prompted the researchers to use fractal thinking to solve content-based image retrieval of related issues. Fractal encoding is used to characterize the image features based on the iterated function system and the collage theorem. The relationship between the recorded image fractal codes to a certain extent reflects the approximation of the structure of the image. So the fractal technology has a great development in the image retrieval. Image features are recorded by cross- scale redundancy in the image, which is the focus of this thesis.In this paper, the basic fractal technology is carried out in the image of encoding and decoding test。Test results as a comparison of the performance of the fast fractal based on the proposed fast fractal encoding. At the same time, the test and analysis of the image retrieval based on the gray histogram is carried out. Three kinds of segmentation strategies based image retrieval efficiency comparison as proposed. Three kinds of segmentation methods are respectively from three different structures to reflect the adaptive of the image. Through the experimental analysis of the image of encoding and search efficiency, the four fork number segmentation algorithm is improved, using a new local code matching block selection. Compared with the related literature, this algorithm increases the number of R blocks in the same layer, and is more prone to find matching blocks in this layer, thus reducing the level of segmentation, and using less fractal code to record the image features and improve the accuracy and timeliness. At the same time, this paper proposes a new HV segmentation strategy, which makes the segmentation algorithm more adaptive and can reflect the cross- scale similarity of internal image. Finally, a fast fractal algorithm based on triangulation is proposed. The algorithm uses an equilateral triangle level of image segmentation, segmentation achieved effective results. In view of the above three segmentation strategies, this paper proposes a new distance formula of R block. And finally gets the similarity between the images by comparing the distance between each R block. Recall and precision effect of these three kinds of segmentation strategies and applications in the test from the formula of the final gallery was much better than the histogram algorithm. As the use of local codebook encoding speed relative to the basic fractal has been greatly improved, achieve practical purposes.