Research On Algorithm Of Content-Based Image Retrieval And Parallelization

Due to the rapid progress of computer technology, especially the digital image processing technology, the number of the pictures is increasing. On one hand, the amount of data in the Internet has attracted more and more people; on the other hand, people find it difficult for them to find the information they really need in a broad array of data-rich image. So reasonable and effective storage management as well as efficient and accurate customer satisfaction retrieve information attracted more and more people’s attention. Faced with massive image data, there is a method to extract image features a single conventional image retrieval methods and to ignore the lack of similarity of the image in the image database, and there is a flaw in processing large-scale data, therefore, we propose the color image features and image texture features to integrate into the comprehensive features, and take advantage of the improved k-means clustering algorithm for image clustering, as well as the use of distributed computing platform based on Hadoop MapReduce parallel algorithm for image retrieval, the main work are as follows:Firstly, in order to improve the accuracy of retrieval, color features are extracted using autocorrelation plot and texture features are extracted using Gabor transform, then they are integrated into the comprehensive features of the image, on this basis, clustering the images in database using the improved K-means clustering algorithm to make similar images to be in a class in the image database to retrieve within a certain range, this paper put forward an algorithm of image retrieval based on color and texture features for clustering(ICTC).Secondly, in order to improve the speed of retrieval, on the distributed parallel computing platform Hadoop, this paper use Map Reduce to implement parallelization of ICTC image retrieval algorithm of image feature extraction, and use MapReduce to implement parallelization ICTC image retrieval algorithm of retrieval, so this way can improve the retrieval speed.Finally, this paper make a large number of relevant experiments and result analyses for image retrieval based on clustering according to Color and Texture features(ICTC) and the parallelization of ICTC image retrieval algorithm on Hadoop platform. Experiments show that compared to color and texture feature clustering algorithm for image retrieval and other conventional algorithms, ICTC has higher accuracy of retrieval; compared to the conventional image retrieval algorithm, the parallelization of ICTC on Hadoop platform has higher speed of retrieval.