Research And Realization On The CT Image Content-based Image Retrieval

Today, under the rapid development of Internet technology and digital equipment, human lifestyle has become more digitized, and various scenes have been retained in the form of image data. People can easily come into contact with a wide range of data everyday, such as photos、meteorological images、frequency、films and medical image. Facing so much image data the traditional text-based retrieval method can no longer meet the normal requirement of people. Then, the content-based image retrieval technology has developed gradually.This paper first introduces the rise of CBIR technology and how it developed, and how the CBMIR technology develop because of the requirement of medical image retrieval. Then the paper discusses the key technologies of CBIR, such as image feature extraction、image feature selection and fusion、similarity measure、image pre-filter、relevance feedback and common search systems. On this basis this paper analyzes the features of medical image and focuses on the CT image and discusses the difficulty of how to apply CBIR into the medical image retrieval. This paper discusses several image retrieval algorithms detailed, such as SIFT matching、Zernike Moment and PseudoZernike Moment, and take these methods into research to search the CT images. Besides, this paper analysis of the experimental results.During the image retrieval stage, this article try to use two attempts, one is using a combination of text and content, which means we will make manual annotations into the images when the quantity of images is not huge. Through this way, we can use key word to remove extraneous images and this will make the image retrieval system spend less time and get higher efficiency. The other is using image edge detection at first. This way can simplify the CT image and make the retrieval easier. The results of the assessment have proved that these two methods are effective.Final, this paper shows a medical image retrieval system which is based on C/S frame. This system is focused on CT images. Users can choose the image retrieval algorithms and see the final search results through the system.