降噪
小波
正多边形
凸函数
解算器
还原(数学)
缩小
规范(哲学)
凸优化
数学
符号
算法
计算机科学
应用数学
数学优化
人工智能
几何学
政治学
法学
算术
作者
Xingwu Liu,Wenhui Lian
出处
期刊:IEEE Transactions on Circuits and Systems Ii-express Briefs
[Institute of Electrical and Electronics Engineers]
日期:2022-12-01
卷期号:69 (12): 5174-5178
标识
DOI:10.1109/tcsii.2022.3197237
摘要
To obtain natural restorations from the noisy images contaminated by speckle noise, this brief presents a novel hybrid non-convex regularizers model for image denoising. The proposed new variational model closely combines the superiorities of non-convex high-order total variation function and $\ell _{0}$ -norm wavelet frame. This combination helps to avoid the staircase artifacts and maintain discontinuities while removing noise. Numerically, by integrating two popular tools: iteratively reweighted $\ell _{1}$ algorithm and variable splitting method, a modified alternating minimization method is adopted to optimize the resulting minimization problem. Finally, compared with several despeckling methods, numerical experiments indicate the competitive performance of our solver in visual improvement and objective measurement.
科研通智能强力驱动
Strongly Powered by AbleSci AI