基于改进的LBF模型的图像分割研究

Research on Image Segmentation Based on an Improved LBF Model

摘要: 局部二进制拟合(LBF)模型已广泛用于图像分割，但其对初始轮廓敏感且收敛缓慢，严重制约了实用性；针对LBF模型过度依赖局部信息、易陷入局部极值、收敛效率低的问题，本文引入CV模型的全局能量项，在局部–全局联合驱动下重新设计能量函数与演化方程，提出一种改进模型(G-LBF)，使曲线可借助图像整体灰度分布快速锁定目标边界。新模型弱化了初始位置依赖，即便起点设置粗糙亦能稳健收敛；与此同时，演化效率显著提高，分割时间大幅缩短，为高精度、高时效的图像分割任务提供了良好方案，在医学影像分析、自动驾驶等领域展现出良好应用潜力。

Abstract: The Local Binary Fitting (LBF) model has been widely used for image segmentation, but its sensitivity to initial contours and slow convergence significantly limit its practical applicability. To address issues such as the LBF model’s overreliance on local information, susceptibility to local optima, and low convergence efficiency, this paper introduces the global energy term from the CV model. Under a joint local-global driving framework, the energy function and evolution equation are redesigned, leading to the proposal of an improved model (G-LBF), which enables the curve to rapidly identify target boundaries by leveraging the global grayscale distribution of the image. The new model reduces dependence on initial contour placement, allowing robust convergence even with coarsely set starting points. Furthermore, it significantly enhances evolution efficiency and substantially shortens segmentation time, offering an effective solution for high-precision and time-sensitive image segmentation tasks. The model demonstrates promising application potential in fields such as medical imaging analysis and autonomous driving.