The Research Of Based On Visual Features Image Classification Retrieval Technology

With the rapid development of electronic science and technology and the increase in Internet speed and universal, people can easily access, exchange and transmission of vast amounts of video and image data. How quickly and accurately to retrieve the images from vast amounts of image data to is an urgent problem. The emergence of image retrieval easy for people to retrieve the image data. So far, the image retrieval techniques can be divided into two broad categories. Text-based image retrieval techniques and content-based image retrieval. Text-based image retrieval techniques, first manually extracting the image keywords used to describe the image contained in the content, and then use the database query technology through the text mode to retrieve the corresponding image. Content-based image retrieval techniques, first extract the image from the image visual features such as color, texture, contour and shape characteristics, then use the retrieved images and the corresponding features of the image database to measure compared to more similar images retrieved . Feature extraction, similarity measure, relevance feedback and performance evaluation is based on content-based image retrieval in a few key technologies. Because text-based image retrieval, there are some limitations, such as manually extracting keywords for image heavy workload of many images is difficult in words an accurate indication of its contents, the human subjectivity led to different people at different times in different occasions on the same image also have a different understanding, and so limited, prompting people to pay more attention to content-based image retrieval techniques.In this paper, based on visual features image retrieval of several key technologies - the user interface, feature extraction, similarity measure, relevance feedback, high dimensional index technical and performance evaluation research, analysis and realized several typical search algorithms, and in presented based on an improved histogram-based image retrieval techniques. The algorithm has two starting points, color space used in accord with human visual characteristics. And the introduction of spatial information to improve the search results of the traditional histogram. Further, through theoretical and experimental analysis shows that image retrieval of image content based on the color and texture features in different distinction between different capacity. Based on those this paper propose an adaptive image retrieval techniques. Finally, on the basis of earlier studies using Delphi 6.0 and SQL Server 2000 designed a scalable image retrieval system for the research provides an experimental platform.