（毕业论文）基于偏微分方程和水平集的彩色图像分割研究.docVIP

摘 要 Gout等人的基于GAC灰度图像选择性分割模型推广应用到彩色图像上，得到基于GAC的彩色图像选择性分割模型。该模型将彩色图像作为整体刻画其边缘并定义其边缘停止函数；利用距离函数体现几何约束，实现选择性分割的目的；采用加法算子分裂算法进行数值求解以提高分割速度。实验结果表明该模型能很好地实现彩色图像选择性分割的目标。 然后，考虑到基于GAC的彩色图像选择性分割模型具有精度低、抗噪性差等缺点，本文考虑将区域信息加入到基于GAC的彩色图像选择性分割模型中，综合利用图像的边缘信息和区域信息，来改善图像分割效果。然后用约束点所围成多边形的内外面积表示几何约束，代替基于GAC模型的距离函数，减少计算量。实验结果表明，相比于基于GAC的彩色图像选择性分割模型，该模型有更好的分割效果，既提高了抗噪性和分割速度，又可以选择性分割特殊形状的物体。 最后，在改进模型的基础上，定义两条水平集曲线，得到双水平集彩色图像分割模型。一条水平集曲线控制选择性分割，另一条水平集曲线控制全局分割。采用约束点围成多边形内外面积表示几何约束进行选择性分割，用最快的速度完成选择性分割。实验结果表明，该模型能实现选择性和全局分割的目的。 关键词：偏微分方程；变分法；水平集方法；彩色图像；选择性分割；活动轮廓 Research on Color Image Segmentation Based on Partial Differential Equation and Level Set Method Yu Yu (Control Science and Engineering) Directed by Professor Li Shurong Abstract Image segmentation is an important task for image processing, which is the foundation of high-level image processing, such as image analysis, image recognition and image understanding. The aim of segmentation is to divide an image into several non-overlapping sub-regions, and pixels in each region share a certain commonality. Segmentation methods based on partial differential equation and level set method, using the thought of dynamic evolution, have the superiority in image segmentation. The basic idea can be summarized as following. Firstly, an energy function about level set curve is defined, and then a partial differential equation controlling the curve evolution can be obtained by variational method, and the final evolution result is the image boundary. Adopting PDE-based image segmentation techniques and level set method; this paper makes some researches on color image as following: Firstly, the GAC-based gray image selective segmentation model of Gout et al. is expended to color image. In this paper, a given color image is treated as a whole and then the edge stopping function is defined. This model describes the geometric constraints by a distance function to selectively segment color images. During numerical implementation, the additiv