Research Related To The Content Based Image Retrieval

With the development of computer and information technology, there are increasingly rich digitalresources on the Internet. Meanwhile, a problem that people are facing is not the lack of digital resources,but how to find what is proper in a large amount of resources. As a result, content-based image retrieval(CBIR) was invented by researchers to solve the problem. But in this field there are still many issues andproblems. Based on the foundation of previous researchers’ work, this study explores and discusses imagefeatures, measures of similarity, color and retrieval methods of textures in content-based image retrieval,and proposes a retrieval algorithm based on color-space distribution according to color features, and ILBP,a improved retrieval algorithm of LBP. There are also discussions of experiments based on the twoproposed retrieval algorithms. The following is what has been done in this study:1. the proposed retrieval algorithm based on color-space distribution according to color featureshas the following advantages over the traditional CBIR: the foundation of the algorithm is people’s visualfeelings, and based on the weighted contrast ratio between pixels and their interfacing pixels the weightedcolor histogram is built. This histogram can reflect the importance of positions of original image pixels.Distribution features of the same color are described by values of dispersion of color-space. The proposedretrieval algorithm based on color-space distribution is designed with the combination of the weightedcolor histogram and the values of dispersion of color-space.2. Based on the feature of the human visual system, a improved retrieval algorithm of LBP isproposed based on the weighted contrast ratio between pixels and their interfacing pixels. The proposedalgorithm has the following advantages over the traditional LBP: the description method of an improvedtexture, this method not only considers the gray mean and the center pixel gray values of local relations,also consider the gray image8neighborhood around the center pixel and pixel value relationship.3. In the present study, experiments based on the two proposed algorithms have been conducted,showing that the performance of them are reliable.