基于Spark的图像检索方法研究

With the rapid development of image processing technology, how to retrieve the image from the mass images efficiently, becomes a problem that needs to be solved quickly. The existing low-level features of images, are lack of semantic information in traditional image retrieval. To solve this problem, a method to extract color information by using twice clustering, was put forward to improve image retrieval. In this method, some semantic information was added to the traditional Hu movement invariants, and multi index based on clustering analysis by shape and the color information compose of multi features. At the same time, compared with serial processing, Spark cluster parallel experiments show that the improved method of search results is more efficiently in dealing with massive images, which indicates that the image retrieval method in practical image retrieval efficiency.