Research On Image Retrieval Technology Bases Color Feature

With the development of multimedia technology, the amount of disorder multimedia information in internet has being increased rapidly. How to locate the useful files in the vast multi-media information is the great technology problem. The content-based image retrieval technology according to visual feature has been carried out as a new method. A content-based image retrieval technique is superior to the traditional text search method. It has been applied in many areas such as multimedia communications, biometrics and intellectual property protection. It has become a research hotspot in recent years and has broad prospects for development.The key technology of content-based image retrieval is feature extraction and expression. The image features most commonly used include color, shape, texture and spatial location. Color is one of the most important features for content-based image retrieval, which is easy to perceive. It has a strong robustness to noise, size, resolution and direction. Color histogram is the common way of extracting color feature. Color histogram can reflect basic color and statistical distribution perfectly. However it can only measure the distribution of color in the overall differences without measuring spatial location information. It is a limitation of Color histogram.In this paper, author proposes an improved method called dominant color matrix based block. By means of color weighted, it can enhance the main color of images and dim secondary color. With the way of block, it can capture the color spatial location relationship in order to enhance the spatial distribution of color histogram. It would improve the method which can only depend on color distribution measuring; furthermore, according to the visual image of regional concern, author proposes to design an UI to select preferential region and adjust the weight coefficient. It can make up the deficiency which the color histogram can not be retrieved. Though resolutions of different pretreatment, it can reduce the computational complexity, improve the operating efficiency and save storage space. Finally I designed the image retrieval prototype system based color feature called MyCBIR. Through several experiments, author analyzes and contrasts different arithmetic in histogram intersection, Euclidean distance and dominant color matrix based block. Also, author proposes the scope of different algorithm's application and verified the improved method's feasibility and the superiority.