ABSTRACT
In the implementation of Snake/CV/RSF model in image segmentation, the boundary stop function defined by the level set method is based on the gray gradient information. It can get better object boundary in high contrast image. But when the image grayscale or contrast is changed, level set segmentation method will drive the evolution curves converge to the pseudo boundary. Because the gray gradient information is sensitive to noise, speed stop function is easy to make the evolution curve stop in a local extreme position or isolated boundaries for those which are not real objects. Phase congruence operator is a function which may detect the same phase points in frequency space to recognize the real boundaries; it is not influenced by the image brightness and contrast changes. So, we can use phase congruence function in active contour model to solve the problem of the traditional active contour curve which easily converges to the pseudo edge in low contrast and gets the correct segmentation for needed objects.
