The Research And Implementation Of Adaptive Image Retrieval System Based On Genetic Algorithm

With the rapid development of computer network and communication technology, multimedia information is increasing. People are increasingly paying attention to how they can quickly retrieve the desired images from the giant image repository and classify. Therefore, the Content-based image retrieval (CBIR) techniques have been proposed, and gradually developed into the core technology research in the field of image retrieval.CBIR is able to use its visual characteristics like color, texture, shape and others to retrieve the desired images quickly. In order to make the image retrieval algorithm based on multi-features have better search results, this paper presents an adaptive image retrieval algorithm based on genetic algorithm. Around the adaptive image retrieval model based on genetic algorithm, the paper proposed block weighted color moment algorithm to express the color features of image on the basis of the research on image retrieval algorithm based on the traditional color moment. Through the comparative analysis of experimental data, which proved that block weighted color moment algorithm has better precision and recall than traditional color moments algorithm. Secondly, on the basis of the traditional image retrieval algorithm based on gray level co-occurrence matrix, this paper proposed Canny edge detection operator to describe the images texture features, to solve the problem of gray level co-occurrence matrix s big calculation. Through the comparative analysis of experimental data, prove the necessity and effectiveness of the use of Canny operator edge detection.In order to change the traditional weights given by hand to find the optimal weights, to make the algorithm more intelligent, this paper proposes the genetic algorithms to find the optimal weights and automatically assign weights, moreover, design and implementation of a image retrieval system, which has the characteristics of human-computer interaction and easy to operate, also can verificate the performance of the algorithm be proposed with the traditional algorithms, Through the comparative analysis of experimental data, which demonstrates the feasibility of the proposed algorithm.