Self-adaptive level set methods combined with geometric active contour

摘要

Level set methods have been extensively used in contour recognition and image segmentation. The traditional level set methods require to initialize the level set function regularly and are strongly dependent on initial contour position thus there are many problems about the methods. Considering those problems, this article proposed a new level set method to draw geometric active contour self-adaptively. Variable weight coefficient is introduced during the curve evolutionary process thus algorithm is independent of contour initial position and evolutionary curves are able to converge to the target boundaries efficiently. What's more, it can recognize the target inner boundaries and the depressed contours thus it is easy to be applied in complicated images. Finally, by computer simulation, the algorithm for recognizing contours under different conditions is proved to be accurate, efficient and reliable.