Research On Multi Feature Fusion Algorithm Of Content-based Image Retrieval

In recent years, with the rapid development of Internet and multimedia information, dataand information explosive in the modern society, How to effectively and quickly access data isthe important research subject in the world. Image data is important in the media of information.It is widely used in the field of digital library, medicine, geography etc. How to rapid andeffective get image data from image database is become the hotspot of scientific research. Basedon content-based image retrieval is produced in order to this direction, the main purposeis provide reliable algorithm to effectively obtain images needed in the database.In this paper, research image retrieval from the image feature extraction, systematicresearch the most basic techniques of image feature extraction, mainly includes the image’scolor, texture and shape features, and improving the original algorithm of extraction to enhancethe availability. The color feature is extracted using the improved model of HSV, because it isconsistent with human vision. Partition algorithm was proposed in the color feature extraction toreduce its dimension. In the extraction of texture feature, using the improved algorithm ofLBP operator, the improvement form of extended circular neighborhood is used in binaryzationof the center pixel, in order to improve the effect of image texture description. In the extractionof shape, considering the problem of the gray image edge contour is not obvious, Image edgeextraction, after enhance the image contrast, and describe the shape feature with8-directioncode chain. This paper compared the influence of different feature extraction techniques to theresults of image retrieval. The image of different feature weight value assignment problem isdiscussed. It is proved by experiment that the improved feature extraction algorithm is effective,the multi feature fusion algorithm is superior to the single feature.At the end of this paper, research image retrieval from the relevance feedback with thedifferent characteristics of image and SVM. Different samples were studied in image retrieval. Itis proved by experiment that the Image retrieval based on feedback feature. It has certainpractical value in image retrieval.