Research Of KLT Feature Points-Based Image Retrieval

With the development of multimedia technique and the widespread of internet, the source of image is expanding, more and more digital images are produced. Image as a rich, intuitive performance of multimedia information has been applied in many aspects. Effective organization, management and retrieval of large image database have become the urgent problem that need to be addressed. Traditional keyword-based or descriptive text retrieval approach has been very difficult to meet people's needs. The content-based image retrieval (CBIR) has become a hot research topic.CBIR uses the visual contents of an image such as color, texture, shape, etc to denote the image. The key technology is the image content feature extraction and matching. This paper investigated the methods of extracting image information based on KLT feature point and space distribution feature after analyzed the key technology of content-based image retrieval and propose a new method based on KLT feature point and space distribution feature. First of all we extract the KLT feature points of image with KLT feature point extraction algorithm. Then, we get the feature of KLT feature point spatial relations histogram. At last in order to get over that space distribution based on KLT feature point only takes into consideration the spatial relations of image while ignores the match of KLT feature point, we propose an integrating the KLT feature point spatial relation and KLT feature point feature method.This paper developed a CBIR simulation according to discussions to the new method. It implemented the new image feature extraction methods and verified its high accuracy and strong applicability by comparing the method with KLT feature point retrieval algorithm and KLT feature point space distribution retrieval algorithm.