改进的LB图像修复模型及其算法  

Improved Lattice-Boltzmann method and algorithm for image inpainting

作　　者：刘作楚[1] 陈翰林[2] 鲜大权[2] 

机构地区：[1]西南科技大学信息工程学院,四川绵阳621010 [2]西南科技大学理学院,四川绵阳621010

出　　处：《计算机应用研究》2013年第10期3181-3184,共4页Application Research of Computers

基　　金：国家自然科学基金资助项目(10971169)

摘　　要：针对格子波尔兹曼(LB)模型模拟图像修复过程,提出了一种基于P-Laplace算子的改进LB图像修复算法。该方法利用P-Laplace算子的非线性各向异性扩散性来填补受损区域,并且忽略了梯度对修复过程的影响。实验结果表明,改进后的LB修复模型,其修复效果比TV和CDD模型好;与LB模型具有相似的修复效果,却比它具有更快的修复速率,适用于苹果采摘机器人视觉系统中对树枝遮挡部分的修复、老照片修复,以及文字移除等领域。To study Lattice-Boltzmann method（LBM） based image inpainting, this paper proposed an image inpainting algo- rithm based on P-Laplace operator. This algorithm made use of the nonlinear anisotropie diffusion of P-Laplace operator to in- paint damaged regions, and what＇ s more,it ignored the effect caused by the gradient. Experimental results show that the pro- posed algorithm has better general performance in convergence speed and inpainting quality. Its objective is to restore the lost information according to around image information, which can be used to restore the shaded part of the branches in image pro- cessing system for aDole harvesting robot, restore old photo, remove text and so on.

关 键 词：图像修补 整体变分 曲率驱动扩散 非线性各向异性扩散 P-LAPLACE算子 格子玻尔兹曼法 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]