PCA-Contourlet Application Research In Image Retrieval

With the development of Internet technology, the way of people’s communication becomes more and more diversity. Image as an effective carrier of information transmission and communication, becoming more and more popular for its rich content of information and easy comprehension.With the development of Internet technology, the way of people’s communication becomes more and more diversity. As an effective carrier of information transmission and communication, Image becoming more and more popular for its rich content of information and easy comprehension.Content-based image retrieval describes itself from a content view of the image. It is a good solution to the defects caused by the traditional marked text-based image retrieval technology. In this paper, from the view point of image texture, studied and implemented the texture feature description methods based on PCA-Contourlet, then applied it into the actual image retrieval project and got good retrieval result. In this paper, the main research work is as follows:(1) Briefly analyzing kinds of image texture features description methods. According to the problem of the ineffective representation in the image multi-resolution and image moving parallel, research and analyze the PCA-Contourlet’s advantages in image texture features description.(2) For the defects of the lack use of low-frequency information during the process of Contourlet transform, using GLCM method to improve the comprehensive of the object texture information description. In allusion to the regional distribution characteristics of the studied contents, referencing sub-block thinking to give the block the characteristic differences in weight distribution.(3) For the problem that the high-dimensional feature vector produced the adverse affect to the retrieve limitation, using the method based on principal component analysis, on the basis of having little effect on the retrieval results, to reduce dimensions of the feature vectors, thus significantly improving the retrieval performance.(4) Design and Implementation CBIR system which based on PCA-Contourlet image texture features expression. Demonstrate the content’s feasibility and applicability in this paper.In this paper, the experimental demonstration of the proposed technologies and solutions for each chapter are given. The results show that the research ideas and technical solutions are effective.