Fast Multi-Object Image Segmentation Algorithm Based on C-V Model

Abstract

Multi-objective image segmentation is a frequently encountered problem. The classical C-V algorithm has the shortage about multi-iterative operations and the computational time is too long to segment the large size image. On the base of analysis upon the relationship between the image size and the number of iterations and time to get the right result, the article proposes a fast image segmentation algorithm based on local C-V active contour model which are based on threshold segmentation and the connected component labeling to segment the large size image with multiple objectives. In the first step, a coarse segmentation is obtained by using the OTSU method, then label and cut the image with the fast non-recursion pixel marking algorithm of connected domains. The segmentation is used as an initial solution in the C-V model. The analysis and experimental results indicate that the improved C-V algorithm can get the right result quickly compared with classical C-V algorithm. It is fast and effective to segment the large size image with multiple objectives.