IMAGE SEGMENTATION BASED ON FRACTIONAL NON-MARKOV POISSON STOCHASTIC PROCESS
Norshaliza Kamaruddin, Nor Aniza Abdullah, Rabha Waell Ibrahim · 2015
This paper introduces a novel region-based active contour model (ACM) with modified fractional non-Markov Poisson stochastic process (MFNMPS) to segment images within intensity inhomogeneity interface in a short time. Utilizing the global ACM, the technique integrates the MFNMPS in its edge enhancement. The MFNMPS with total energy is employed to enhance rapid contour movement within an enhanced of edge in the image texture. The fractional Poisson stochastic process with its counting function could classify the inhomogeneous object in a region in providing a smooth homogeneous region for segmentation with low computational cost. The MFNMPS with global energy allows the contour development to transfer toward the object founded on the preserved edges in providing improved segmentation. The conforming EulerLagrange is applied within the level set framework to minimize the energy. This study practices the proposed process to segment images with intensity inhomogeneity in segmenting several synthetic and medical images. With improved speed, the proposed method more accurately segments medical images compared with other baseline approaches.