Research On Image Retrieval Algorithm Based On Fractal Technology

With the rapid growth of the amount of multimedia information,the demand for processing multimedia information accurately and rapidly has become more and more urgent.Corresponding computer information processing technologies have been more widely used.Content-based image retrieval technology occupies the field of image analysis Important position.Since the nineties,image compression algorithms based on fractal technology have developed rapidly and gradually become mature.The mature application of fractal technology in image compression has led researchers to turn their attention to the problems related to content-based image retrieval using fractal technology.Therefore,on the basis of studying the structural similarity between images,this paper improves the existing fractal methods,and realizes the image coding,decoding and fractal-based image retrieval algorithms.The improved algorithm is tested and analyzed through experiments.First of all,this article has carried on the research and the test to the commonly used picture code,the decoding technology and the image retrieval technology at present.Among them,the fractal technology based on fixed block is tested in the test library,and the result of image coding and decoding is taken as the reference of the improved algorithm in this paper.The gray histogram retrieval algorithm is tested on the standard ORL library,The comparison of fast fractal retrieval algorithm proposed by the paper.Then,this paper improves the existing image segmentation methods,and uses quadtree segmentation and HV segmentation methods.A new local codebook selection strategy is proposed.Different methods are studied from the aspects of segmentation method and local codebook selection strategy The degree of adaptation to the image.The improved image segmentation method makes it easier for R block to find matching D blocks in the current layer.Meanwhile,the selection strategy of local codebook greatly reduces the number of D blocks in D pool and thus reduces the image encoding time.After that,according to the different distance formulas proposed in this paper,the distance weights of approximate D blocks in different directions are weighted according to the selected features of the fractal codes to reflect the similarity distance measure more accurately.Finally,the improved algorithm in this paper is tested on the standard gallery and living gallery.The experimental results show that the use of local codebook can greatly improve the image encoding speed relative to the fixed block fractal.To face the image to be retrieved,The encoding algorithm can be used to generate the fractal code to match the similarity and meet the requirements of real-time retrieval.By using the improved distance formula,the precision of the test library is obviously better than the gray histogram algorithm.