摘要
提出变量分离和交替最小化相结合解决l1正则优化问题,并用于非纹理图像的修复。基于变量分离技术,该算法将目标函数的l1成分和l2成分解耦,l1正则优化问题简化为一系列非约束优化问题。除了交替最小化迭代地求解这些非约束优化问题,还引入投影法加快和简化求解过程。实验在有噪声和无噪声的情况下,用提出的算法对信息丢失30%的图像进行修复。实验结果表明:该算法可有效解决包括图像修复在内的一系列图像复原问题;与某些同类算法相比,在修复速度和修复效果方面均具有优势。
An algorithm,which used variable splitting and alternate minimization to solve the l1 regularized optimization problem,was proposed and applied to image inpainting.Based on the variable splitting,the proposed algorithm decoupled the l1 and l2 portions of the objective function,which reduced the l1 regularized problem to a sequence of unconstrained optimization problems.The alternate minimization was used to solve these unconstrained optimization problems with the projection algorithm for accelerating and simplifying the resolving procedure.With and without the noise,the experiments were carried out on the images with 30% pixels missing.And the experimental results demonstrate that the proposed algorithm can solve a series of image restoration problems including the image inpainting.Compared with other similar algorithms,it shows competitive speed and inpainting results.
出处
《计算机应用》
CSCD
北大核心
2011年第8期2206-2209,共4页
journal of Computer Applications
基金
国家自然科学基金面上项目(61070090)
国家自然科学基金青年科学基金资助项目(61003270)
淮北师范大学校青年科研项目(700442)
关键词
l1正则优化问题
图像修复
变量分离
交替最小化
投影算法
l1 regularized optimization problem
image inpainting
variable splitting
alternate minimization
projection algorithm