Robust image segmentation model based on binary level set
Wenqi Zhao
TL;DR
This work tackles robust segmentation of images with intensity inhomogeneity and noise by adopting a binary level-set formulation augmented with a GL operator to reduce reinitialization and enhance noise resilience. A bias-field corrected intensity model, I = b c, is integrated into the binary level-set framework, yielding an energy functional that combines data fidelity for inside/outside regions with a gradient-based regularization and a binarity-enforcing term. The optimization proceeds via a three-step splitting operator with alternating updates for $c_1$, $c_2$, $b$, and $\phi$, including implicit, FFT-based, and projection steps to maintain binary values. Empirical results using Dice and JS metrics on bias-field images demonstrate improved robustness to intensity inhomogeneity and noise, highlighting potential benefits for medical and real-world image segmentation.
Abstract
In order to improve the robustness of traditional image segmentation models to noise, this paper models the illumination term in intensity inhomogeneity images. Additionally, to enhance the model's robustness to noisy images, we incorporate the binary level set model into the proposed model. Compared to the traditional level set, the binary level set eliminates the need for continuous reinitialization. Moreover, by introducing the variational operator GL, our model demonstrates better capability in segmenting noisy images. Finally, we employ the three-step splitting operator method for solving, and the effectiveness of the proposed model is demonstrated on various images.