Image Retrieval Based On Color And Texture

With the rapid development of network and multimedia technology, content-basedimage retrieval technology is becoming a hot research. It is mainly to extract the imageof the underlying characteristics such as color, texture, shape, etc, then retrieve thesimilar images from the database which is similar to the query image, at last return theretrieval results to the user. After the study we find that the integration of multi-featurehas good retrieval performance.In this paper, it mainly introduces the image’s color and texture features, featureextraction techniques, common color models, color quantization, feature similaritymetrics and performance evaluation criteria. Then it studies the integration of color andtexture image retrieval method, experiments show that it effectively improve theefficiency of image retrieval. The main research work and innovation of this thesis canbe stated as follows.1. Retrieval technology based on color image is summarized, as well as color spaceand color quantization. In this paper, a new HSV quantization is proposed whileextracting the color features, that is two uniform division of the H component, anon-uniform division of S and V components, at last we combine with the average ofthe two quantifiable results as a color histogram.This method takes into account thecontinuity of the color range, experiments show that it has good retrieval performance.2. Retrieval technology based on texture image is summarized, the definition andclassification are introduced and analyzed. Then the statistical analysis, structuralanalysis, model analysis and spectrum analysis are researched and summarized. Animproved direction local binary pattern is proposed. It considers not only therelationship between the local center and the local pixel, but also the relationshipbetween the local gray mean and the center pixel. Experiments show that the algorithmhas a very good retrieval performance on the texture image retrieval, and it has moreobvious advantages on the color image retrieval, and shortens the retrieval time.3.Retrieval technology based on color and texture is introduced and researched,through experimental analysis, the proposed combination of color and texture has betterretrieval performance.