摘要
全变差(TV)图像复原正则化模型一般由正则项和保真项两部分构成。针对该模型容易形成使边缘平滑和产生阶梯效应的问题,在修改正则项后提出了一种新的图像复原模型。改模型利用交替方向乘子算法来优化其求解模型,即利用辅助变量把全变差复原问题转化为一个等价的无约束优化问题,基于交替方向乘子迭代将无约束优化问题分解为几个子问题,再根据子问题的特点,利用阈值法对问题进行优化求解。实验结果表明,所提出的新模型能有效保护图像边缘并抑制阶梯效应,明显地提高了图像的质量;与其他正则化图像复原模型相比,其具有较高的信噪比,较小的相对误差和较好的图像恢复效果。
Total Variation （TV） image restoration regularization model generally consists of two parts：regularization term and fidelity term. In view of the problem that the model could easily make edges smooth and produce the stair effect, a new image restoration model with modified mgularization term has been proposed, which can be optimized with alternating direction multiplier algorithm to achieve its solution model. The total variation restoration problem can be transformed into an equivalent unconstrained optimization problem by utili- zing auxiliary variable and then the unconstrained optimization problem can be disassemble into such several sub-problems that each sub -problem can be optimized and solved with threshold value method according to the peculiarities of the sub-problem structure. Experi- mental results show that this new model proposed has significantly improved the image quality since it can protect image edges and re- strain staircase effect, and that compared with other regularization image restoration models,it helps achieve higher signal to noise ratio, smaller relative error and better effect of image restoration.
出处
《计算机技术与发展》
2017年第4期60-63,68,共5页
Computer Technology and Development
基金
国家自然科学基金资助项目(11461037)
昆明理工大学人才基金(2008-72)
关键词
全变差
图像复原
正则化
阶梯效应
交替迭代算法
total variation
image restoration
regularization
stair effect
alternative and iterative algorithm
