选择性计算的快速非局部均值图像去噪

针对非局部均值(NLM)图像去噪算法度量像素间的相似性计算强度高的问题,提出了一种选择性计算的快速NLM去噪方法。在图像块像素灰度值向量空间距离计算时,利用L2范数逐次消元法,只需在图像积分图上通过少量加法运算即可剔除大量相似性低的像素点,有效地减少计算强度。根据图像空间相关性强的特点,提出了基于patch测地线距离的动态调整搜索区域的方法。实验结果表明,与其他经典算法相比,该方法获得了较好的加速,也提升了NLM算法的去噪性能。

Fast Nonlocal Means Image Denoising Algorithm Using Selective Calculation

A fast nonlocal means (NLM) image denoising method with selective calculation is proposed to solve the problem that the computational cost of similarity weights is high. By using L2 Norm successive elimination, a large number of pixels of low similarity van be rejected through a small amount of additive operations on integral image, and the massive calculation on measuring similarity can be effectively reduced. According to spatial coherence in the image domain, an approach for adaptive search area based on patch geodesic distance is proposed. Experimental results demonstrate that the proposed method, compared with the state-of-the-art algorithms, can not only accelerate the nonlocal means algorithm, but also elevate the image quality.