Image Retrieval Based On The Feature Of Color And Texture

At present, Content-Based Image Retrieval (CBIR) is becoming a hot research topic. Breaking through the limitations of traditional retrieval technology based on the text, it directly analyzes image content and extracts image features, such as the color, texture, shape and spatial relations of the image, and so on, then builds index and carries on the retrieval using these content features.This thesis firstly pointed out the important meaning of content-based image retrieval technology to the science research and industry production based on the background of problem studied, and analyzes the current status and developing trend of image retrieval. And then the basic principles, features and problems of image retrieval were summarized. The algorithms and process of image retrieval based on color, texture and fusing both feature were researched thoroughly and deeply.As for color-based image retrieval, an approach of image retrieval based on dominant color of partition was proposed. It designed the new overlapped sub-blocks and individually weighted. Meanwhile, reduced quadratic form distance method was applied to the algorithm. The algorithm makes use of the image color information, and adds color space information simultaneity. As for texture-based image retrieval, this thesis presented the method of extracting texture feature based on improved wavelet transformation, through the study of wavelet analysis. The algorithm carried on the limited re-decomposition to the high, medium and low-frequency parts of the image and avoided the nonessential decomposition process. After decomposing, the image wad districted and individually weighted. This method increased the middle part texture weights of the image. The experiment results indicate that this algorithm more accurate expresses the image texture information and improves the retrieval efficiency.To overcome the partiality of retrieval based on a single feature, comprehensive feature image retrieval was researched. The thesis combined approach of image retrieval based on dominant color of partition and the method of extracting texture feature based on improved wavelet transformation, considering comprehensively both color and texture features. Through the image retrieval experiment, better image retrieval performance can be achieved by combing two kinds of features.Based on the above algorithms, the paper design and develop a retrieval system based on color and texture, which can retrieval image by color, texture or fusing both features. In the system, a simple relevance feedback was applied, automatically analyze the best useful feature of expressing inquire target from consumer feedback information. With plentiful experiments, it is proved that the adaptive capacity and retrieval efficiency can be enhanced by using this method.