Global and Local Signed Pressure Force Functions Active Contour Model Based on Entropy

Abstract

Image segmentation is considered as a challenge in MRI images, synthetic and real images, because of intensity inhomogeneity which alters the final result of segmentation. Paper presents a novel method which deals with intensity inhomogeneous images. The proposed method introduces global and local fitting image energy function, where global fitting image is based on entropy and, moreover, presents global and local signed pressure function for stabling the gradient descent flow that is solving the energy function. Local signed pressure function has been established by multiplying entropy with local image difference and global signed pressure has been calculated by global image difference. Entropy used for estimating the bias field, global fitting term focused on homogeneous regions as well as provided the robustness to initialization of contour and local fitting term useful to detect objects with intensity inhomogeneity. Experimental result shows that in the presence of entropy, this method gives the superior performance concern to accuracy, time, and robust to initialization.