Research On Generalized M-j Set Retrieval Method Based On Color Feature

With the rapid development of network and information technology, the amount of imagedata increase sharply. How to find the required image information becomes more difficult. Thetraditional method of image retrieval is to search according to the text label marked on image,which is named text-based image retrieval. The marked text (text label) needs needs adding inmanual way, which is really a labor consuming work. Furthermore, it is difficult to express allthe contents only by the scripts. This problem becomes the bottleneck with the huge demands ofimage searching. So, the content-based image retrieval is introduced.The content-based image retrieval is based on the automatic extraction of images of internalvisual features (color, texture, shape and other features), calculate the distance between theimage features and features of example images，then the degree of similarity between the imagescan be expressed by the distant similarity. Compared to the text-based image retrieval, the mainidea of this approach is the automatic extraction of visual features among the objective images,rather than relying on the additional text marked, which can save time and human resources andalso have greatly improved the overall efficiency of the retrieval.This paper focuses on the generalized M-J sets retrieval method by color features, and thebasic idea is to use the more common image retrieval technology, combined with the intrinsiccharacteristics of the generalized M-J sets and put forward certain original retrieval method.Because of the complex color scheme of the generalized M-J sets, the retrieval of single color isdifficult to achieve our requirements. In this research, we reviewed and studied many cases onsingle color-based image retrieval methods. Then the image database at the base of our previouswork, the FDS system, is set up as our research objects for the color feature-based imageretrieval method. After many experiments on the FDS fractal images, the result showed thefeasibility of our algorithm. In this research, we also integrated the characteristics of a variety ofcolors and proposed the corresponding retrieval algorithms for the generalized M-J sets. Themainly innovative points are as follows:(1) Establish a dedicated color feature of the generalized M-J sets coloration, according to the coloration method, carry out color quantization to the generalized M-J sets. On this basis,the generalized M-J set’s color features are extracted.(2) On the images produced by the generalized M-J sets in our previous project, FDS(Fractal Design System), we introduced a novel retrieval algorithm based on the color characterin the base of point (1). The algorithm retrieves fractal images from different levels. Throughanalyzing the color scheme of the generalized M-J sets, three aspects of the color histogram,color moment and color aggregate vector, are used to describe the characteristics of each color.The integration of three features can reduce the drawbacks of every single one, which will behelpful for the introduction of more effective retrieval algorithms.(3) Proposing workable generalized M-J sets image retrieval algorithm based on colorfeature, and developing a system applying generalized M-J sets retrieval method.This study will promote the application of the generalized M-J sets of image retrieval basedon color features. The results showed by the experiments are promising.