Image Retrieval Based On The Feature Of Texture And Color

With the development of computer and communication, image processing are applied widely to many fields. In order to retrieve image well, contended-based image retrieval comes into being and becomes a hot question in image processing field. We focus on the image retrieval in the feature of texture and integration.Gray histogram is a simple method describing image feature, but it doesn't include the spacial information of gray. The gray concurrence matrix can describe the texture feature better. As describing the texture feature full-scale, we combine the information of gray and grads, forming a gray-grads concurrence matrix, statistic and getting 15 texture descriptor. Before getting the gray-grads concurrence, we divide image into pieces which can add more spatial information. Wavelet-transform can decompose the image into high-frequency parts and low-frequency parts. High-frequency parts figure the image detail and low-frequency parts figure smooth part. We can use the coefficient of high-frequency to describe the texture feature. In paper, we use tree-structured wavelet transform to decompose the image. When computing the texture feature, we divide the high-frequency image into three parts, highlighting the middle part of the image. Experiments show that this method can reflect texture and obtain good retrieval effects.Integrated retrieval can integrate the advantage of different feature. we introduce a method in color feature and experiment the integrated method using texture and color feature. Based on the research above, an experimental system of image retrieval is designed. It is used to check feature extraction methods mentioned above. It is an opened structure, and users can add their feature extraction methods into it by some rules anytime.