非局部均值（NLM）滤波图像去噪技术

非局部均值（NLM）滤波是一种广泛应用于图像去噪的方法。相比传统的局部均值滤波，NLM滤波器能够更好地保护图像的细节和边缘信息。在NLM滤波中，像素的权重不仅取决于其邻域像素，还取决于与整个图像中其他像素的相似度。因此，NLM滤波器能够更好地去除噪声，同时保留图像的细节和边缘信息。
下面是一个简单的Matlab源码实现非局部均值滤波的示例：
```matlab
function [denoised_image] = nlmeans_filter(noisy_image, h, window_size)
% NLM FILTER: NON-LOCAL MEAN FILTERING
% This function applies the non-local means filter to a noisy image.
% Inputs:
% noisy_image: the input noisy image
% h: the standard deviation of the Gaussian kernel
% window_size: the size of the local window
% Outputs:
% denoised_image: the denoised image
% Convert the image to grayscale if it is not already
if size(noisy_image, 3) == 3
noisy_image = rgb2gray(noisy_image);
end
% Preallocate memory for the denoised image
denoised_image = zeros(size(noisy_image));
% Apply the non-local means filter to each pixel in the image
for i = 1:size(noisy_image, 1)
for j = 1:size(noisy_image, 2)
% Extract a local window around the current pixel
local_window = noisy_image(max(1, i - floor(window_size/2)): min(size(noisy_image, 1), i + floor(window_size/2)),
max(1, j - floor(window_size/2)): min(size(noisy_image, 2), j + floor(window_size/2)), :);
% Compute the weighted average of the pixels in the local window
weighted_sum = 0;
total_weight = 0;
for k = 1:window_size^2
[row, col] = ind2sub([size(local_window, 1), size(local_window, 2)], k);
similarity = exp(-0.5((noisy_image(i, j) - local_window(row, col))/h).^2);
weighted_sum = weighted_sum + local_window(row, col) similarity;
total_weight = total_weight + similarity;
end
% Assign the denoised pixel value to the current pixel in the denoised image
denoised_image(i, j) = weighted_sum / total_weight;
end
end
end