Research Of Snake Contours Extraction Algorithm Based On Medical Image

As one research branch in the field of computer vision, medical image processing is an important application of computer graphics and digital image processing in the biomedical engineering. 3D reconstruction and visualization based on contours of sequent medical images are hotspots in current research, and widely used in diagnostic, surgery planning, simulating, teaching in anatomy and medical simulate training. Therefore, research for the contour-extraction algorithm, which is the preprocessing of 3D reconstruction, is both of high theoretic meaning and of extensive application perspective. The active contour model (snake) is a considerably potent instrument in medical image segmentation and contour extraction, owing to its remarkably active property. In this paper we study some crucial technical problems, with witch we are usually confronted when it comes to applying the Snake model to extracting contours of those issues or organs in medical images.Conventional image smoothing methods usually destroy image features such as edges, lines and textures when they eliminate noises from the original images, which is a misfortune for the later contour extract. To preserve the useful features, an anisotropic diffusion algorithm based on partial deferential equation is studied in this paper, and an improved method is presented. It is proved by experiments that the novel method is able to suppress the isolated noises effectively as well as to preserve those important features.The kernel of applying the Snake model to extracting contours from medical images concerns how to define a reasonable energy function as well as which strategy should be used to guarantee the convergence to the global minimal energy in short time. Based on the geometrical model of Snake, we propose the A* search procedure based snake algorithm to settle the two critical problems. First, a weighted average method is used to construct a reasonable graph for the later search, reflecting the snake energy of each pixel in the original image. Second, A* search algorithm is applied to finding out the contour between two specified points. Because the A* algorithm searches for the globally best resolution with heuristic knowledge, the fast convergence to the global minimal energy is guaranteed. Applying the novel snake algorithm to actual contour extraction leads to satisfying result, because it can not only simplify snake's initialization to decrease its sensitivity to its initial position, but also converge to the global minimal energy quickly. Experiments show that our snake contour extraction algorithm has accurate and robust performance.After having extracted contours from the sequent medical images, 3D reconstruction from medical images is under our consideration. To get the tetrahedral mesh of organs and issues, we first trianglate the contours on each section to get 2D triangle facets, and then link facets between two adjacent sections to form tetrahedral. This is an open topic calling for more intensive research.