一种半离散的非线性图像增强方法

提出了一种半离散的非线性扩散方法。借助SIDEs不连续右端的基本思想,将整个框架在弹簧体模型下进行考虑,给出了一些好的性质,并对该扩散方法提出了两种数值求解方案:显式方案和可加算子分裂(AOS)的迭代格式。在显式方案中,通过一个特殊的矩阵分解能大大减少滤波过程的运算量;在AOS方案中,对时间步长没有限制,从而可以选择较大的时间步长,提高了运算效率。实验表明,该扩散进程既有很好的平滑效果,又能增强图像边缘,且运算时间较短。

Semi-discrete nonlinear algorithm for image enhancement

A semi-discrete nonlinear algorithm is proposed.Following by the idea of discontinuous right-hand sides in SIDEs,the whole framework is studied based on the spring-mass model,and some good properties are given.Furthermore,two schemes are presented: explicit scheme and additive operator splitting(AOS) scheme.In explicit scheme,a matrix with symmetry and vanishing row sums is proved to have a special decomposition,which can reduce the complexity during the diffusion process.In AOS scheme,the time step can be set very large to improve the model's time efficiency.Experimental results show that better smoothing is obtained and image edge is enhanced well.