Study On Medical Image Segmentation Method Based On Active Contour Model

The validity of medical image segmentation directly affects the subsequent image analysis, recognition and the high accuracy of auxiliary diagnosis of the subsequent image. Therefore, the medical image segmentation has a great significance in the medical image processing. The traditional image segmentation methods existed many limitations and can not meet the high qulity of the image segmentation. The Parametric Active Model (also named Snake) derives the evolution of the contour based on the prior model and image data fitting functions. It can obtain the boundary of target region through continuous envolution. The traditional Snake model has got a wide application since it was put forward.However the conventional Snake model can not segment the image with concave areas well and it has a limitation in obtainning a wide capture range. This thesis has proposed an improved the Dynamic Directional Gradient Vector Model combined with wavelet transform. The algorithm firstly decomposes the image by the multi-analysis of the wavelet, and then results in the improved DDGVF segmentation under each layer of the wavelet decomposition of the image, so that it can acquire a more accuracy target contour at last. In order to verify the validity of the improved ailgorithm, the synthetic images and the real complex medical images has both carried on the conventional algorithms and the impoved algorithm respectively in MATLAB simulation experiments. The simulation results have proved that the proposed algorithm solve the limitation of the Snake model. Furthermore, it can better segment the depression area of the target object image and get a wider capture range than the conventional Snake model. That is to say the improved dynamic directional gradient vector flow combined with wavelet tranform algorithm is a more accurate medical segmentation method.The Geometric Active Contour Model solves the segmentation based on the level set method, and it has received more researches and more application fields since it has been proposed. Based the consideration of the diversity of medical images, intensity inhomogeneity and the existence of artifacts and noise, the traditional geometric active contour model such as C-V model based on the single global regional information and the sementation method based on the single boundary information can not accquire an accurate and stable segmentation result. Thus this paper has introduced a region scalable fitting function, and has proposed a new mixed model based on both the edge and region information. The algorithm first designs an appropriate velocity function and then combines with a region scalable fitting energy function. By combining the region and boundary information together, it can make the curve envolution to the boundary of the target area and can obtain an accurate stable segmentation result ultimately. Through the contrast of the MATLAB experimental results, it has proved that the modified algorithm can effectively reduce the influence of noise and artifacts and improve the segmenting quantity of intensity inhomogeneity images. In addition it has a better segmentation effect and a faste segmentation speed. That is to say, it is more suitable for image segmentation field. Finally, the thesis summaries the disadvantages of traditional algorithm and the advantages of improved algorithm and then prospects future works.