摘要
对于脉冲噪音和模糊图像,最常见的恢复方法是全变分加1范数,即TV/L1模型。但是,对于高噪音水平的情形,TVL1模型的恢复效果不太好。为解决上述问题,本文提出一种新的模型,即在TV/L1模型基础上,加上一个由反正切函数构造的线性的修正项。模型求解采用交替方向法。数值实验验证了本文所提出的新方法的有效性,尤其对于高噪音情形,去除脉冲噪音的效果明显优于TV/L1模型。
The total variation (TV) regularization term plus L1 norm, denoted by TV/L1 model, is widely used to the problem of image restoration where the observed images are corrupted by blur and impulse noise. However, TV/L1 model may produce a poor recovery solution, especially for high noise levels. In order to overcome the problem, we propose new modification of TVL1 (MTV/L1) which a linear correction term, constructed by an arc-tangent function, is added. Alternating di-rection method of multipliers (ADMM) is presented to solve the TV/L1 and MTV/L1 models. Nu-merical experiments verify that our proposed approach outperforms TV/L1 in terms of sig-nal-to-noise ratio (SNR) values and visual quality, especially for high noise levels.
出处
《计算机科学与应用》
2017年第2期124-128,共5页
Computer Science and Application
