Research On The Methods For Breast X-Ray Image Preprocessing

Breast cancer is a common malignant tumor of the female in cancer. Mammography radiography technique is the preferred choice for early detection and early diagnosis of the breast cancer. Since the breast X-ray projection imaging with low pixel and unclear, we need a technique which can improve image quality by according to the characteristics of breast X-ray image, to make the unclear picture’s feature more obvious. Then use the appropriate image segmentation technique, to assist doctors with more effective observation and analyzing image. Help doctors to identify the breast masses easier.Against the problem with mammography projection imaging, this paper presents a new method for mammogram image enhancement based on wavelet fusion. Anymore, gives an improved live-wire interactive segmentation algorithm to edge division of suspicious masses under doctors controlling. The data-sources in experiments come from the Mammographic Image Analysis Society mini-database and clinical mammography images. The result indicates that the methods have good effect.In this paper, the following work with mammography image preprocessing:(1) This paper presents a new method for mammogram image enhancement based on wavelet fusion. The algorithm based on the image preprocessing with histogram equalization method and median filter, an efficient fusion method which adopt the average rule to low-frequency and the variance rule to high-frequency, was proposed to realize the enhancement for mammography image. Contrast with multiple images in experiment, this paper puts forward information entropy, mean gradient and CII that evaluate the enhancement algorithms. The experimental results show that this method can improve the image contrast, and can restrain noise and keep the edge details, which getting a good effect on mammography image enhancement.(2) Enhanced images for image segmentation, this paper describes an improved live-wire interactive segmentation algorithm to edge division of suspicious masses, using Canny operator instead of LoG operator, reconstructing the cost function, and redistributing weights by according to the characteristics of mammography image itself. Meanwhile set qualification to make the shortest path searching more effective. The experimental results indicate that this algorithm can reduce noise response, at the same time, accurately locate actual edge to guarantee edge’s continuity. To extract the edge with mass more accurately which get the better segmentation for breast X-ray masses. Against the problem with mammography image having more nodes, set qualification for making the shortest path searching to reduce the unnecessary nodes searching, can reduce the amount of computing to improve efficiency.