Research On The Application Of Texture Analysis And Nonlinear Dimension Reduction Method To Image Retrieval

With the development in multimedia and network technology, image plays an important role in diary life. It is an urgent task to efficiently manage the image database which is growing quickly. Content-based image retrieval is the key technique to solve the problem that is how to retrieve useful information within enormous amount of images.Extracting low level visual features from images is the first step of image retreival, and that texture feature is an important low level visual feature. In this paper, besides using existed texture feature, I proposed some new texture features: multi-scale complexity feature multi-scale fractal dimensional number feature and 2D Hilbert spectrum feature. In order to solve the border problem , we proposed a broder-symmetry arithmetic which is based on k-mean clustering. At the same time, I used some existed color feature. Experimental results showed that these existed and proposed features were suitable for image retrieval.In Content-Based Image Retrieval , the spatial complexity of high eigenvector's storage and the operation complexity of computing their similarity will be much higher. In this paper, we used nonlinear dimensionality reduction to reduce both spatial and operation complexity. The experiments showed that it can get good results.At the same time,we used relevance feedback and machine learning used in image retrieval. K-NN, BP neural network and support vector machine classifiers were used in experiments. Image retrieval accuracy raised by using relevance feedback and machine learning.A general-purpose image retrieval experimental platform was also developed. Reseachers can use this platform not only to carry out all kinds of image retrieval experiments without any coding, but also do academic intercommunion conveniently. This platform is very useful for image retrieval study.