An active contour method using harmonic mean

Abstract

Active Contour method has been shown very effective in detecting the contour of region(s)-of-interest(ROI) and is widely used in image processing and computer vision. In this work, we aim to improve the performance of Zhang's method in detecting boundary of ROIs. Specifically, we will generalize the CV energy functional and give a new special case. The new energy functional penalize the approximation error (of the original image by a constant) weaker than the CV energy functional, which can better preserve the subtle difference between the intensity of ROIs and that of the background, thus can effectively segment images, especially images with low contrast. The resulted two-phase constant approximation is the harmonic mean instead of the arithmetic mean. Based on this, we improve Zhang's active contour method by using the harmonic mean. We apply the proposed method on synthetic and real images and the segmentation results show that the proposed method is robust to noise and intensity contrast. Additionally, the proposed method is less sensitive than Zhang's method to parameter selection.