Segmentation Of Breast X-ray Image Based On Wavelet Analysis And Genetic Algorithm

The X-ray mammography is a low-noise and low-contrast image,so the digital X-ray mammography in the early detection of breast cancer is a challenging task. Therefore the need for effective noise suppression image to get reliable results. Furthermore, due to the damage partially embedded and hidden in a variety of breast soft tissue structures in different densities, so the damaged part of the segmentation is a difficult task. In order to give the radiologist to accurately diagnose breast X-ray images to help, a detailed study of the pre-processing of the image (image denoising and image enhancement) and suspected tumor segmentation method. The main work is done as follows:Study the existing segmentation algorithms and image pre-processing methods to analyze the wavelet (multi-scale geometric analysis) analysis theory and application. Curvelet transform based multi-scale properties, including not only the scale and location, which also includes a very detailed direction, the removal of image noise and artifacts depending on the nature of its finest scale and most coarse scale, mass and denoised image background, low contrast, blurred edges clarity of image denoising performed after Curletvel transform a series of low and high frequency sub-band coefficients of low frequency sub-band coefficient nonlinear enhancement to achieve the purpose of image enhancement.Proposed segmentation method based on wavelet transform and genetic algorithm combining.This method combines wavelet transform algorithm and genetic algorithm This method combines wavelet transform algorithm and genetic algorithm.. first,discrete wavelet transform to decompose the image, extracting artwork in different dimensions and details of the components of the approximate weight, Then the image histogram to approximate wavelet decomposition after treatment with the genetic algorithm, the number of threshold and threshold size, Finally, the resulting threshold extension to the original space, the expanded use of a threshold for breast X-ray image segmentation.In 30 breast X-ray images containing lesions as the research object, comparing the algorithm and traditional proposed segmentation algorithm simulation the segmentation evaluation data in Matlab environment. The results show that:Compared with the traditional breast segmentation algorithm, the algorithm edge higher positioning accuracy, good continuity edge, complete tumor segmentation and detection of weak edges are also very effective to successfully breast X-ray images of the segmentation process.