S. Phumeechanya, C. Pluempitiwiriyawej, and S. Thongvigitmanee (Thailand)
Active contours, deformable models, image segmentation,level set method, snakes
In this paper, we propose a novel active contour method using a local region-based force with adaptive-length search lines. Our method uses the intensity values of the pixels along a set of search lines that are perpendicular to the contour front. These search lines are to tell the contour front which direct to move toward the object boundary. More over, each of these search lines is not of a predetermined fixed length. Instead, its length increases until it can determine, using the local regional information, to which direction the correspondingcontour pixel should move. We have compared the performance of our method to several other edge- and region-based active contours. The results show that our method provides desired segmentation results on some particular images where other methods may fail. Not only is our method robust to noise, and able to reach into a deep concave shape, it also has a large capturing range and performs well in segmenting heterogeneous objects.
Important Links:
Go Back
