Robust Image Segmentation Algorithm Based on Local Correntropy-Based Generalized Loss Function

Image segmentation has been an essential area of research, and traditional image segmentation is one of the most vital research branches. Improving the robustness of image segmentation models is still a critical topic because of the influence of noise and the presence of intensity inhomogeneities such as varied light levels in the actual images. In this paper, we propose a new robust image segmentation algorithm based on the correntropy criterion and generalized loss function based on the level set model. The local correntropy criterion and generalized loss function makes the new proposed model insensitive to outliers, and also has good robustness to noise distortion and interference. The proposed method is evaluated by comparison experiments on real and synthetic images, and it finally shows that compared with other traditional segmentation methods, the proposed method has further improved in terms of robustness and accuracy of segmentation, and the effect of initialized contours on the segmentation effect is further reduced.