Directionlets And Its Application In Mammography Enhancement

Breast cancer is one of the most common malignant tumor diseases among women. It is also the second leading cause of death in women. Mammography enhancement is one of the key steps in breast cancer CAD (Computer Aided Diagnosis) system, through which the visibility of mammography could be effectively improved and more reliable diagnosis information are available to the doctor.This paper is based on the study of multiscale geometric analysis and its application in image processing. We have studied the Directionlet Transform, Nonsubsampled Directionlet Transform, Compressed Sensing and their application in mammogram enhancement. And series of new enhancement methods are proposed in this paper for enhancing the natural images and mammograms. The main works of the thesis are summarized as follows: firstly, based on the study of Directionlets, we propose an effective enhancement algorithm through combining Directionlets with Generalized Gaussian Mixture Model. The method effectively improves the contrast and visibility of subtle abnormities. Secondly, to preserve more information of the original image, the Nonsubsampled Directionlet transform is implemented by removing the sampling operation in the Directionlet transform. Then we combined the NSDT with nonlinear method, which is very effective for natural images while not work well on mammograms. For further improving the visual quality of mammograms, the NSDT and GGMM enhancement method is proposed, which actually highlights the subtle edges of abnormal regions. Finally, we studied the compressed sensing technology and its application in mammogram enhancement. The image compressed sensing based on Directionlet transform has been realized first. And then a new image enhancement method based on NSDT and CS has been proposed. The new method can effectively enhance the abnormities on mammography, and meanwhile suppress the negative impaction of background and noises.Experimental results show that the proposed series of enhancement methods are effective on both natural image and mammography, which have broad application prospects.