Research And Implementation On Content-based Medical Image Retrieval

With the improvement of image-based clinical technique, large amounts of digital medical images are produced everyday. These images greatly enrich researchers'consult, research and teaching, but it is also a difficult problem for the researchers to manage the great number of images. Traditionnal text-based medical image retrieval has a lot of disadvantages, which depends on personal idea too much and can not reflect the primitive characteristic of the image, so this method is out of date. In this situation, content-based medical image retrieval develops rapidly.At first, this paper summarizes the key technologies in this field. Based on adequately reviewes current methods of retrieval, then considers the special characters of medical images, this paper focus on researches a few algorithms which are suitable for medical. In terms of gray feature, with regard to the deficiency of global gray histogram, a novel method combining local gray histogram and gray spatial distribution entropy was presented. In terms of texture feature, specific to the flaw of gray co-occurrence matrix, an image retrieval algorithm first taking a image preprocessing, then extracting the block weight co-occurrence matrix feature was introduced. In terms of shape feature, aiming at the defect of the Canny edge detection algorithm, an new algorithm combining improved Canny edge detection and Zernike moment was presented. At last a method for image retrieval using combined all these features was proposed, furthermore, relevant feedback was used to gain high retrieval efficiency.According to all the algorithms above, this paper realizes a prototype system for content-based medical image retrieval. The system uses lots of CT images to test the efficiency of the algorithms. Results show that all the algorithms had good performance.