Pruned non‐local means

Abstract

In Non-Local Means (NLM), each pixel is denoised by performing a weighted averaging of its neighboring pixels, where the weights are computed using image patches. We demonstrate that the denoising performance of NLM can be improved by pruning the neighboring pixels, namely, by rejecting neighboring pixels whose weights are below a certain threshold λ. While pruning can potentially reduce pixel averaging in uniform-intensity regions, we demonstrate that there is generally an overall improvement in the denoising performance. In particular, the improvement comes from pixels situated close to edges and corners. The success of the proposed method strongly depends on the choice of the global threshold λ, which in turn depends on the noise level and the image characteristics. We show how Stein's unbiased estimator of the mean-squared error can be used to optimally tune λ, at a marginal computational overhead. We present some representative denoising results to demonstrate the superior performance of the proposed method over NLM and its variants.

Mapping Heterogeneous Buried Archaeological Features Using Multisensor Data from Unmanned Aerial Vehicles

“…The technique found to be the most useful was non-local means denoising, a filter which protects image texture. The average value of pixels NL u (p) is evaluated by windowing across the entire image, first searching the image for pixels that resemble those requiring denoising [67][68][69] and output is achieved by performing a weighted averaging of neighboring pixels, where the weights are computed using image patches. It is given by:…”

Noise bias compensation for tone mapped noisy image using prior knowledge

“…This class of filters replaces pixel-wise calculation of the distance with patch-wise one. Reports on the NLM filter have been actively studied, such as improvement of denoising performance [11][12][13], processing speed [14][15][16], combination of NLM filter, and another method [17,18].…”

Region characteristics-based fusion of spatial and transform domain image denoising methods

“…There are several critical issues regarding the NLM algorithm such as patch size, search region size, weight calculation, smoothing parameter, and computational complexity, which are still being researched [12]. Several variants of the NLM algorithm have been developed to handle these issues effectively [13][14][15][16][17][18][19][20]. Wavelet transform has become a popular and efficient tool for image denoising due to its various properties such as energy compaction, orthogonality, low complexity, and linearity [21].…”