高阶全变差正则化彩色图像去马赛克  

Color Image Demosaic Model Based on High-order Total Variation Regularization

作　　者：刘铭丽 王希云[1] LIU Ming-li;WANG Xi-yun(College of Applied Sciences,Taiyuan University of Science and Technology,Taiyuan 030024,China)

机构地区：[1]太原科技大学应用科学学院

出　　处：《太原科技大学学报》2019年第6期493-497,共5页Journal of Taiyuan University of Science and Technology

基　　金：山西省自然科学基金(2008011013);山西省“131”领军人才工程项目

摘　　要：针对全变差(total variation,TV)正则化方法会在恢复出来的图像中产生阶梯效应,提出了一种基于四阶偏微分方程的高阶全变差正则化彩色图像去马赛克模型,此模型在ROF模型的基础上,用四阶范数代替TV范数,消除了ROF模型所产生的阶梯效应。实验结果表明,该模型能取得较好的图像复原效果。Because of total variation regularization method(total variation abbreviated as TV)can produce staircasing phenomenon in recovery of image,a high-order model based on a fourth-order partial differential equation(PDE)is proposed for color image demosaicking which improved the ROF model.The new method introduces fourth-order norm instead of TV norm to overcome the staircasing in reconstructions from ROF mode.The experimental results show that the model can achieve better image restoration effect.

关 键 词：去马赛克 全变分 四阶偏微分方程 阶梯效应 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]