Font Size: a A A

Research On Content-based Medical Image Retrieval

Posted on:2013-07-23Degree:MasterType:Thesis
Country:ChinaCandidate:G H CaiFull Text:PDF
GTID:2298330467978118Subject:Computer technology
Abstract/Summary:PDF Full Text Request
With the digital medical imaging equipment such as digital X-ray, CT, MRI and ultrasound are widely used in clinical medicine, large amounts of digital medical images are produced everyday.These images greatly enrich and make it convenient for the medical people and the scientist to work. But it is a problem the medical people and the scientist have to face that how to manage and orgnize these images effectively. So the CBMIR has become a hot topic in recent decade.In this paper, we studied deeply the algrithm of medical image retrieval based on the gray, shape and texture features after reading a lot of related journals, papers and books. We proposed a new kind of two-level algorithm that the first-level retrieval algorithm is based on the gary and shape features and the second-level retrieval algorithm is based on the texture feature of the medical image. The main content of this paper is:1. For the gray feature of the image, the traditional gray histogram only reflects the statistical characteristics and does not reflect the gray space information. So we propose a new method that we divide the image into several rings and extract the statistical characteristics of the histogram of each ring, and then we calculate the space distribution of entropy among the rings which can reflect the gray structural information;2. For the shape feature,, this paper proposed a new method based on the wavelet modulus maxima combined with improved Canny operator to overcome the problem of ignoring low-frequency information and the discontinuous boundary. Method of wavelet modulus maxima is applied to high-frequency part of wavelet decomposition image and Low-frequency part is based on improved Canny operator. Then the two boundaries of image are fused. And then the shape feature was extracted from the obtained boundary image and shape-density histogram was constructed. Lastly the gray feature and the shape feature are fused and the first-level medical image retrieval was executed according to similar calculation.3. For the second-level retrieval algrithm which is based on the texture feature, we propose a new method of improved Gray Level Co-occurrence Matrix(GLCM). In this method, we apply the Gradient Phase Mutual Information(GP-MI) combined with the method of masked image to overcome the shortcoming that the traditional GLCM is impacted greatly by the rotation angle and the background region. The method of GP-MI is applied to compute the angle between two images and the method of masked image is applied to remove the background of the image. After these two steps, we divide the image into several blocks equally and compute the GLCM of every blocks. Then we sum the GLCM of every blocks by different weights as the final texture feature. Lastly medical image retrieval was executed according to the similarity calculation.4. We designed a CBMIR software system based on the algorithm in this paper.We made a lot of simulation experiments and the results indicate that the algorithms proposed in this paper have a promising effect. And the test showed that this CBMIR software system had a strong application.
Keywords/Search Tags:medical image retrieval, gray histogram, wavelet modulus maxima, gray levelco-occurrence matrix, multi-level and multi-features
PDF Full Text Request
Related items