Research And System Implementation Of Sketch-Based Image Retrieval

With the fast development of multimedia technology, computer technology and communication technology, multimedia images exploding, content-based image retrieval technology has aroused widespread concern of scholars both at home and abroad and they have achieved significant research results.In recent years, smart phones, tablet computers and other touch-screen devices is becoming an indispensable tool in our lives, the way of human-computer interaction changes a lot, touch is becoming the mainstream input method. All ages, all levels of painting people can easily draw objects on the touch screen coming to mind to form hand-drawn sketches, and then through the sketches to find images with similar shapes in a large photo gallery. As the sketch describing the peripheral contour information of the object is more aligned with people’s subjective visual experience of the object in the process of perception. Therefore, sketch-based image retrieval technology did a lot of attention.The dissertation focuses on the related image retrieval core technology, and designs a sketch-based retrieval prototype system, the dissertation’s main work is listed below:Firstly, some algorithm of vision feature extraction, similarity measurement and performance evaluation criteria are introduced in detail, as well as some ways of the sketch preprocessing.Secondly, a new retrieval algorithm which combines improved color histogram and edge histogram is proposed. For the defect of missing color spatial distribution information of global color histogram method, this algorithm extracts the whole image’s global color histogram at first, then uses the improved weighted block color histogram to extract local color histogram, combines with the edge histogram to extract image’s shape contour feature, allocates different weight synthesis improving search algorithm performance.Thirdly, on the basis of shape context, a dynamic shape context retrieval algorithm based on statistics is proposed. This algorithm dynamically selects the number of feature points, and statistics the number of feature points of target shape in different sub-block forming contour feature point histogram, and finally uses dynamic programming algorithm for the similarity measure, the experimental effect is good.Finally, a sketch-based image retrieval prototype system is designed, the purpose of the system is to test the proposed algorithm in this dissertation.