Research On Content-Based Image Recognition And Retrieval Technology And System Realization

With the rapid development of computer, multimedia network and automation technology, image using is becoming more and more widespread in various areas, the management and the recognition and retrieval of image information resources are becoming more and more important day by day, and content-based image recognition and retrieval has already become one of the most active research areas in the latest few years. In order to improve inquiry accuracy rate of image recognition and retrieval system, an improved algorithm based on color feature is proposed and single features are synthesized into global feature to recognize and retrieval images;The way of parting and weighing is used in the improved algorithm based on color feature in order to embody space information of color. The improved algorithm of color feature and the classical coocurrent matrix algorithm of texture feature are integrated into the global method. Global features are matched through weighing color feature and texture feature. Weight value can be adjusted by users according to characteristics of images in image DB. It can be also adjusted through relative feedback in order to obtain the best effect of image recognition and retrieval;Essential image preprocesses before feature extracting are researched. Systematic and thorough researches on recognition and retrieval methods as for color and texture feature of image contents and application system realization are conducted. And an improved algorithm based on color feature is proposed and the effects of image recognition and retrieval are improved theoretically. Finally, in order to come off the limitation of image recognition and retrieval based on color feature purely, an integrated method is proposed with color feature and texture feature;Finally, in order to verify the feasibility and inquiry accuracy of the above algorithms, firstly, an image DB which includes 130 concerned images (these images are all preprocessed) is established; secondly, image features which are extracted by system are saved in Accessdatabase which has been already founded beforehand; thirdly, on Microsoft Visual Studio.NET development platform, prototype application system of these algorithms is designed and developed using VC++ including four function modules of image preprocessing, feature extracting and entering DB, recognition and matching and demonstration. Finally, verification of these methods and experimental analyses of this prototype application system are conducted. And experimental results indicate that,inquiry accuracy degree of this integrated method proposed in the topic is higher than image recognition and retrieval methods based on single features.