Abstract: As to the problems of difficult and inaccurate in noise image segmentation, and the existing models are poor in robustness to complex noise environment. This paper proposes the noise image segmentation algorithm based on improved energy functional model, which combines anisotropic diffusion equation and level set method. Through the improvement of energy functional it can segment noise images accurately and quickly. Firstly, non-convex functional is introduced into the energy functional model, and it is proved that there is no global minimum in the energy functional model. Object outline is detected by energy functional model based on non-convex functional. Secondly, smooth images are obtained by anisotropic?diffusion model, the improved energy function model is obtained by combining the smooth images and level set model. Finally, The optimal position of the level set evolution of the foreground target is obtained by solving the minimum value in the energy functional. The experimental results show that the proposed algorithm has higher segmentation accuracy and robust for noise image than other similar models.

Key words: image segmentation, noise, energy functional, level set, anisotropic diffusion equation

摘要： 针对噪声图像的分割难，分割不准确，以及现有模型无法适应多种噪声环境的问题，提出了一种基于改进的能量泛函模型的噪声图像分割算法，该算法结合各向异性扩散方程和灰度水平集算法，通过对能量泛函的改进实现对噪声图像的准确、快速分割。将非凸泛函引入能量泛函模型，并通过证明不存在全局最小值，利用获得的能量泛函模型得到光滑的目标图像边界。将各向异性扩散模型得到的光滑图像与水平集模型相结合，得到改进的能量泛函模型。通过求解在能量泛函的最小值，得到前景目标的水平集演化的最佳位置。该算法与同类模型的对比实验表明该模型对噪声图像具有较高的分割精度和鲁棒性。

关键词: 图像分割, 噪声, 能量泛函, 水平集, 各向异性扩散方程