Research And Realization Of CBIR System Basted On Domain Color And Its Distribution

With the rapid development of the techniques of multimedia and Internet, the application of image becomes wider and wider. And CBIR has become one of the focuses of research recently. Joint with techniques of image processing, pattern recognition, image comprehension and database application, it is a new potential technique in the field of researches and applications.The most important method of image retrieval is its color-based technique, In this paper, several key problems in applying color to retrieval image-image preprocessing, color implication, color characteristics extraction and color-based similarity measurement- are discussed. And at last a prototype system of image retrieval is realized by Visual C++ 6.0.In this paper, the main works are as follows.â—† To discuss the process of image preprocessing. Color image denoising and graying are mainly expounded. The color image denoising algorithm based on rough set theory and color image graying method based on the similarity of dominant color are presented.â—† To show color characteristics of an image, the HSV color space, which is suitable to the visual characteristic of human, is utilized in this paper, and a new color table for each image is established by color cluster in this color space.â—† An improved method of color histogram-The Domain Color Similarity Histogram (DCSH)-is presented to improve the precision.â—† In order to describe the domain color's spatial characteristics, the algorithm of Domain Color Similarity Statistically Differential Histogram (DCSDH) is brought forward.â—† To express the characteristics of the a color image, the method of DCSH and DCDSH are combined together.â—† Finally, the retrieval result is analyzed, and a conclusion is given out.