This is machine translation

Translated by Microsoft
Mouseover text to see original. Click the button below to return to the English version of the page.

Note: This page has been translated by MathWorks. Click here to see
To view all translated materials including this page, select Country from the country navigator on the bottom of this page.

Remove Noise from Color Image Using Pretrained Neural Network

This example shows how to remove Gaussian noise from an RGB image. Split the image into separate color channels, then denoise each channel using a pretrained denoising neural network, DnCNN.

Read a color image into the workspace and convert the data to double. Display the pristine color image.

pristineRGB = imread('lighthouse.png');
pristineRGB = im2double(pristineRGB);
title('Pristine Image')

Add zero-mean Gaussian white noise with a variance of 0.01 to the image. imnoise adds noise to each color channel independently. Display the noisy color image.

noisyRGB = imnoise(pristineRGB,'gaussian',0,0.01);
title('Noisy Image')

Split the noisy RGB image into its individual color channels.

noisyR = noisyRGB(:,:,1);
noisyG = noisyRGB(:,:,2);
noisyB = noisyRGB(:,:,3);

Load the pretrained DnCNN network.

net = denoisingNetwork('dncnn');

Use the DnCNN network to remove noise from each color channel.

denoisedR = denoiseImage(noisyR,net);
denoisedG = denoiseImage(noisyG,net);
denoisedB = denoiseImage(noisyB,net);

Recombine the denoised color channels to form the denoised RGB image. Display the denoised color image.

denoisedRGB = cat(3,denoisedR,denoisedG,denoisedB);
title('Denoised Image')

Calculate the peak signal-to-noise ratio (PSNR) for the noisy and denoised images. A larger PSNR indicates that noise has a smaller relative signal, and is associated with higher image quality.

noisyPSNR = psnr(noisyRGB,pristineRGB);
fprintf('\n The PSNR value of the noisy image is %0.4f.',noisyPSNR);
 The PSNR value of the noisy image is 20.6395.
denoisedPSNR = psnr(denoisedRGB,pristineRGB);
fprintf('\n The PSNR value of the denoised image is %0.4f.',denoisedPSNR);
 The PSNR value of the denoised image is 29.6857.

Calculate the structural similarity (SSIM) index for the noisy and denoised images. An SSIM index close to 1 indicates good agreement with the reference image, and higher image quality.

noisySSIM = ssim(noisyRGB,pristineRGB);
fprintf('\n The SSIM value of the noisy image is %0.4f.',noisySSIM);
 The SSIM value of the noisy image is 0.7393.
denoisedSSIM = ssim(denoisedRGB,pristineRGB);
fprintf('\n The SSIM value of the denoised image is %0.4f.',denoisedSSIM);
 The SSIM value of the denoised image is 0.9507.

In practice, image color channels frequently have correlated noise. To remove correlated image noise, first convert the RGB image to a color space with a luminance channel, such as the L*a*b* color space. Remove noise on the luminance channel only, then convert the denoised image back to the RGB color space.

See Also

| | | | | |

Related Topics