The Design And Implementation Of Content Based Image Search System

With the rapid development of high speed Internet and multimedia technology, the searching demand of image information is becoming more and more. How to find the needed image information in the massive network images becomes a valuable research topic.Content-based image retrieval technology analyzes images according to the image visual features such as color, texture, shape and the relationship of spatial position, and conducts the image retrieval by extracting image features to establish the database of image feature vector. In this paper, we first introduced the related extraction technologies of image global and local feature, improved a color feature extraction method based on block histogram. Experiments showed that this method has better retrieval performance than global histogram and traditional block histogram. Compared the different local feature extraction technologies of the differences in time and the number of feature points through experiments, and chose the best SURF local image feature extraction algorithm. Then compared and analyzed the differences of different algorithms for local feature matching in speed and performance, and chose the optimal k-d tree algorithm used to match the feature points of the image. The index of the feature space was established by the random k-d tree in order to improve the search efficiency.Based on the requirement analysis, this paper designed all modules of the image search system including image acquisition, image feature extraction and matching, image indexing and user interface, and implemented an Internet image search system based on the improved block color histogram and the local SURF feature. Experiments showed that the image search system based on integrated features effectively avoided the limitations of single feature search, improved the recall ratio and precision of search results. Finally, the test data set was used to test the system, and the performance of the system was further optimized. The test results showed that the system has realized the real-time search for large scale image database, the performance and efficiency of the system has reached the expected goals with good image retrieval performance.