Image Retrieval And Application Based On Color And Shape Features

With the coming of the information explosion era, large-capacity storage devices and high-performance electronic products came into being. Multimedia and network technology have spread quickly; massive images, video, data have grown at an amazing speed, and people's needs for searching and retrieving messages have become increasingly urgent. However, the traditional text-based image retrieval (Text-Based Image Retrieval, TBIR) can no longer meet the satisfaction of people for the target data. Instead, the content-based image retrieval (Content-Based Image Retrieval, CBIR) is proposed. For its efficient prosecution rate and query accuracy degree, it is concerned more and more, and has become the hot spot of solutions for the image data problems at this moment.The object of Content-Based Image Retrieval is the image itself. The method for indexing is extracting features of the bottom image (such as color, shape, texture, etc.). Then compare the distance between the query criteria and these characteristics to determine the degree of similarity of two images. Finally, give the image data with the similarity up to the set threshold to the query user.The paper, based on the study and research about recognition computing and computer vision, combines with individual knowledge and understanding for content-based image retrieval, designs and implements a simulation experiment environment of road traffic sign recognition. With the robustness of color feature extraction in space and the robustness of shape feature extraction in the displacement transformation, a multi-feature matching algorithm which is combined of color features and shape features is used. In the respect of color feature, some improvements are made to the traditional color histogram method, and a new color histogram method based on main colors is proposed. By combining the major color retrieving method and the color histogram computing, two quick screenings are carried out, thereby the scope of the search is narrowed and the retrieval efficiency is improved. With Fourier shape descriptors adopted, an improved contour-based description method is proposed. Because the tangential angle of contours (curvature) is highlighted and factors such as complex coordinates and center distance are ignored, within a reasonable range, the accuracy is lowered appropriately and the query speed is improved significantly. However, a single feature extraction method, after all, have some limitations of its own existence, it is unable to play fully to its all advantages. Experiments show that, if we can combine the features of the color and shape together, adjust the weighting coefficientλbetween the two, ultimately determine the value ofλthe recall ratio and precision ratio will reach a relatively higher level.The organization structure of this paper is as follows:Chapter One, I mainly introduce the objective, meaning and research background of this topic and state of development at home and abroad in this field; Chapter Two, I mainly introduce the related knowledge and method of image retrieval technology based on content, including the retrieval levels of the image content, low-level visual features description method, feature matching technology and performance evaluation criteria and so on; Chapter Three, I mainly introduce image retrieval technology based on color, including the selection of color space, color quantization, color histogram, color entropy, color moments and so on. I also improve a new color histogram method according to practical application; Chapter Four, I mainly introduce image retrieval technology based on shape, including image segmentation, image retrieval based on outline, image retrieval based on area, and so on. I also improve a new outline description method according to practical application; Chapter Five, I introduce the multi-structure integrated technology, adjusting method of image feature weights and finally establish the weight; Chapter Six, I process the simulation experiment according to the weight established in Chapter Five, and finally make the conclusion, prove that the arithmetic after optimizing is more correct an efficient; Chapter Seven, I summarize the whole paper, propose the application area and feasibility.