自适应的L1-L2范数正则化图像去噪方法  

Adaptive L1-L2 Norm Regularization Method for Image Denoising

作　　者：豆泽阳 毕翔[1] 曹宝杰[1] 

机构地区：[1]中国传媒大学理工学部理学院,北京100024

出　　处：《中国传媒大学学报（自然科学版）》2016年第1期50-55,共6页Journal of Communication University of China：Science and Technology

摘　　要：提出了一种自适应的L1-L2范数正则化图像去噪方法。相比传统的L1范数正则化与L2范数正则化,新方法有效消除了阶梯效应,同时较好的保持了图像边缘信息。为了提高计算效率,将Split Bregman算法框架应用到提出的模型中,有效的提升了收敛速率并减少了计算时间。实验结果与分析验证了L1-L2范数正则化模型在图像去噪效果与计算效率的有效性。An adaptive L1- L2 norm regularization method for image denoising is proposed.Compared to the traditional L1 norm regularization and L2 norm regularization,the proposed method removes the staircase effect effectively and maintains the edge information of the image.In order to improve the computation efficiency,the Split Bregman algorithm is applied to the proposed model,which effectively improves the convergence rate and reduces the computation time.Experimental results and analysis verify the effectiveness of the L1- L2 norm regularization model in the image denoising effect and the computational efficiency.

关 键 词：图像去噪 自适应 L1-L2正则化 SPLIT Bregman迭代 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]