机构地区: 深圳大学数学与计算科学学院
出 处: 《电子学报》 2012年第4期769-772,761,共5页
摘 要: 本文给出一种新的图像多尺度表示算法.首先,应用OSV模型得到图像的单尺度分解;其次,针对上一步的信息亏损,引入不同的单调尺度参数,迭代OSV变分模型,从而为图像的不同特征提供一种非线性的分级自适应表达式.同时,本文也给出有关新算法的离散格式.数值实验表明,与已有的Nezzar算法相比,新算法的多尺度分解效果更佳. We propose a new multiscale image representation based on OSV model.We obtain with OSV model,in which we give single scale decomposition.Then we give a hierarchical adaptive representation for the different features in images by iterating OSV model for different monotone scale parameters,applying the decomposition to the residual information of the previous step.The result multiscale decomposition is nonlinear.We discuss the numerical discretization of the new method.The numerical results show an excellent decomposition effect and significant improvement over Nezzar method.