The Research For The Algorithm Of Image Retrieval Based On Combining Color And Texture Feature

With the rapid development of information industry, the amount of media data is increasing. The image is the important carrier to obtain information and data is more and more big, people need to study how to search available data from the large amourifof images, content-based image retrieval (CBIR) is based on using image low-level features for image retrieval. After describing the previous image retrieval research, It analyzes the shortcomings of single feature, and proposes the combination of color and texture features of images retrieval method, and uses image preprocessing and relevant feedback mechanism to improve image retrieval conditions to achieve better retrieval results. Firstly, extracting color and texture features, then calculating color and texture features, and then output the results.In feature extraction.it describes several color models and color extraction methods, and analyses the advantages and disadvantages of various methods, and using color histogram to extract color features. It analyses the wavelet transform method and gray level co-occurrence matrix method of the advantages and disadvantages. Gray level co-occurrence matrix method is more adaptable, considering the spatial location of the texture, so it selects of gray co-occurrence matrix method; in the part of similarity calculation, histogram correlation uses good characteristics of the color histogram, and the calculation is simple, so the color features using histogram correlation distance to calculate the similarity; Euclidean distance is used widely and convenient to calculate, and texture features using Euclidean distance similarity calculation.In this paper, it has studied the image preprocessing mechanism and the related feedback mechanism to improve the image retrieval effect. Before the image retrieval, it pre-processes the image, and selects the image area, removes the interference factor, and the image retrieval is more specific. Using the feedback mechanism, checking whether the results meet the requirements of the user, it can change the retrieval conditions, change the retrieval method to get better retrieval results.Finally, in order to verify that the image retrieval method, using Corel image data sets containing interference image to carried out image retrieval experiments; using image retrieval evaluation precision (precision) and recall (recall) to evaluate the retrieval results. Use The relevance feedback mechanism to change the retrieval conditions, Using the image pre-processing mechanism and retrieval experiments in order to obtain a better retrieval effect.