基于二维局部均值分解的自适应保真项全变分图像滤噪方法

摘要

For effective edge feature preservation and good denoising result, this paper proposes a new adap-tive fidelity term total variation image denoising algorithm based on bidimensional local mean decomposi-tion and local high-frequency energy. Firstly, an image is decomposed into different scale parts from low frequency to high frequency through using bidimensional local mean decomposition algorithm; secondly, the highest frequency component can be used to calculate the local energy function, and then the adaptive fidel-ity term is achieved; finally, image denoising can be performed by minimizing the energy functional. This paper implements a parameter sensitive analysis for adaptive fidelity term total variation image denoising algorithm to verify its high uncertainty. In comparison with other image denoising algorithms, our algorithm presents a better result in edge feature preservation and noise removing under strong noise. Our algorithm solves some problems existing in other denoising algorithms, e.g., staircase effect, speckle effect and poor denoising results near the edge. The experimental results approve that our algorithm has a better robustness and more faster processing in comparison with the adaptive fidelity term total variation image denosing algorithm.%为了在滤除图像噪声的过程中既保留图像的边缘细节，又对噪声有良好的滤除效果，提出一种基于二维局部均值分解和局部高频能量的自适应保真项全变分图像滤噪算法。首先采用二维局部均值分解算法自适应地将图像分解成从高频到低频不同尺度的成分；然后将其中最高频的成分用于计算局部能量函数，求得自适应保真项参数；最后通过求解最小化能量泛函实现图像噪声滤除。实验结果表明，该算法能较好地保留图像的细节边缘，即使在强噪声下也能较好地对图像平滑区域实现滤噪，解决了其他算法在保留边缘的同时产生的阶梯效应、斑点效应以及边缘附近噪声滤除效果差等问题；且相比于自适应保真项全变分图像滤噪等典型算法，具有更好的鲁棒性与更快的处理速度。