基于图像局部方向特性的自适应全变分去噪模型  

Adaptive directional total variation denoising model based on image local direction

作　　者：唐玲[1] 陈明举[1] 

机构地区：[1]四川理工学院自动化与电子信息学院,四川自贡643000

出　　处：《液晶与显示》2016年第5期477-483,共7页Chinese Journal of Liquid Crystals and Displays

基　　金：人工智能四川省重点实验室开放基金项目(No.2012RYY08;No.2013RYY02);四川省教育厅科研项目(No.13ZB0138;No.14ZB0211);四川理工学院校级科研基金项目(No.2012KY13)~~

摘　　要：针对全变分模型(total variation,TV)以图像的梯度信息作为去噪的尺度参数,未考虑图像局部纹理的方向性的缺点,提出了一种基于图像局部方向特性的自适应全变分去噪模型(Adaptive directional total variation,ADTV),并推导出该模型的迭代数值求解过程。在该模型中,首先,计算出图像局部方向的角度矩阵。然后,构造与图像纹理方向一致的椭圆区域代替TV模型的圆形区域。最后,通过优化最小化算法迭代求解以获得去噪后图像。通过对比实验证明,本文提出的模型取得了更高的峰值信噪比,去噪过程中更好地增强了图像的细节信息。In order to overcome the disadvantages that total variation model based on the image gradi- ent information can＇t consider image＇s local direction and remove noise effectively, an adaptive direc tional total variation model based on image＇s local direction is proposed,and the iterative procedure of model is deduced. First, the image local direction is calculated in this new model. Second, an ellipse coinciding with the dominant direction of the local image is used to replace the unit ball in the total variation model. Finally, the denoised image is obtained by using iteration to optimize the minimum total variation. The comparing experimental results show that the proposed model can get higher peak signal to noise ratio and enhance the image details, and it has good performance in image denoising.

关 键 词：图像去噪 全变分 优化最小化算法 图像局部方向 

分 类 号：TN911.73[电子电信—通信与信息系统]