A fusion-based approach to image segmentation using ROI and multi-phase level set model

A new approach to image segmentation is proposed, which is based on regions of interest and improved multi-phase C-V model. Different from the traditional C-V model, the improved multi-phase level set model include more information of the image, such as image intensity information, area, circumference, and the image gradient information. In addition, adding the edge indicator and penalty term to the model to improve the efficiency of the approach. The use of the image gradient information can overcome the inaccurate edge localization defects in multi-objective image segmentation. The edge indicator is used to keep the boundary information better in the evolution process and the penalty energy can effectively eliminate the re-initialization of the SDF and therefore accelerate the evolution speed of the level set. Meanwhile, in order to fasten the contour's convergent speed and enable the avoidance of trapping, we make a coarse segmentation by using FCM clustering which is utilized to initialize contour location. After the evolution process, extract the targets according to the proposed ROI algorithm. Experimental results show that the proposed approach has advantages in accuracy in comparison with the classical multi-phase C-V model on medical image segmentation.