线性化临近点算法及其在视频监督中的应用

Linearized Proximal Point Algorithm and Its Application in Video Surveillance

基于参数化临近点算法和线性化加速技术,构造了一种线性化临近点算法求解二分块可分离凸规划模型,从而将子问题转化为临近算子的形式便于求解.利用变分不等式框架分析了算法的全局收敛性和次线性收敛速度.应用所提算法对视频监督中的一类图像处理问题进行求解,相关仿真实验测试了该线性化加速算法的数值效率.

Based on the parameterized proximal point algorithm and a linearized acceleration technique,we construct a linearized proximal point algorithm for solving the two-block separable convex programming model,which could transform the involved subproblems into proximal mappings to be solved easily.The global convergence and sub-linear convergence rate of the new algorithm is analyzed by using the framework of variational inequality.The proposed algorithm is finally applied to an image processing problem in video surveillance,and related simulation experiments test the performance of this linearized accelerated algorithm.