This article proposes a level set approach to segment images with N regions by using a single level set function. Many works use several level set fronts or hierarchical approach to solve this problem. We propose a general formulation of the level set propagation function based on the knowledge for two regions. A modified likelihood function is proposed, and each background probability is maximized. A single propagation function is achieved from N logarithmic components. Experiments are performed on synthetic images with normal probability and computed tomography images of patients with hemorrhagic stroke. Our approach is compared with other ones known in the literature, and the level set was superior in 10 metrics out of 13 evaluated, with an accuracy of 99.67% and FSIM 93.96%. These results confirm the effectiveness of the proposed method.

本文提出了一种通过使用单个级别集功能来分割具有N个区域的图像的级别集方法。许多作品使用多个级别的集合前沿或分层方法来解决此问题。我们基于两个区域的知识提出了水平集传播函数的一般表述。提出了一种改进的似然函数，并使每个背景概率最大化。从N获得单个传播函数对数成分。对出血性中风患者的具有正常概率的合成图像和计算机断层扫描图像进行实验。我们的方法与文献中已知的其他方法进行了比较，在13个评估指标中，有10个指标的水平集更好，准确度为99.67％，FSIM为93.96％。这些结果证实了所提出方法的有效性。