LBF model energy function is non-convex with level set function, leading to extreme sensitivity of LBF model segmentation final results with level set function initialization. By convex LBF model energy function, the paper presents a gobal LBF (GLBF) model. GLBF model is convex with respect to level set function, so that by arbitrary initializing level set function, a global optimum result is obtained. Moreover it is unnecessary for the model to re-initialize level set function as signed distance function, so that it greatly speeds up the computation efficiency. Segmentation results with unevenly-gray medical images explains that GLBF model is insensitive with level set function initialization, so that it is superior over the traditional LBF model and state-of-the-art representative LIF model.%LBF模型的能量函数对于水平集函数是非凸的,从而导致应用LBF模型分割的最终结果对水平集函数的初始化非常敏感.通过凸化LBF模型的能量函数,提出一种全局的LBF模型(GLBF).该模型针对水平集函数是凸的,从而可以通过任意初始化水平集函数得到全局最优解.此外,该模型不必重新初始化水平集函数为符号距离函数,从而极大地提高运算效率.对灰度不均匀医学图像的分割结果表明,GLBF模型对水平集函数的初始化不敏感,优于传统的LBF模型以及目前具有代表性的LIF模型.
展开▼