High-precision inhomogeneous image segmentation based on adaptive parameter level set method

Abstract

In the field of image segmentation, there are still some technical problems, such as the segmentation accuracy of the image with intensity inhomogeneity is not high enough. Level set method has the advantages of solid theoretical foundation and stable algorithm for segmentation of small sample data, so far it has played a significant role in image segmentation. In this paper, with the level set method for the research framework, we present a new approach to segment noise and weak boundary images that is supported by an adaptive velocity correction function. Based on the properties analysis of noise and edge information, this correction function is replaced the original constant to largely improve the segmentation accuracy. In addition, on the basis of some theoretical analysis, we use the edge stop function constructed by entropy to effectively improve the accuracy of image segmentation. Finally, a large number of challenging image experiments show that the proposed segmentation model performs better than other classical models in segmentation accuracy.

Keywords

• image segmentation
• level set method
• velocity correction function
• intensity inhomogeneity
• entropy