denoiseImage

Denoise image using deep neural network

Syntax

B = denoiseImage(A,net)

Description

B = denoiseImage(A,net) removes noise from noisy image A using a denoising deep neural network specified by net.

This function requires Deep Learning Toolbox™.

Examples

Remove Image Noise Using Pretrained Neural Network

Load the pretrained denoising convolutional neural network, "DnCNN".

net = denoisingNetwork("DnCNN");

Load a grayscale image into the workspace, then create a noisy version of the image.

I = imread("cameraman.tif");
noisyI = imnoise(I,"gaussian",0,0.01);

Display the two images as a montage.

montage({I,noisyI})
title("Original Image (Left) and Noisy Image (Right)")

Remove noise from the noisy image, then display the result.

denoisedI = denoiseImage(noisyI,net);
imshow(denoisedI)
title("Denoised Image")

Input Arguments

A — Noisy image

2-D image | stack of 2-D images

Noisy image, specified as a single 2-D image or a stack of 2-D images. A can be:

• A 2-D grayscale image with size m-by-n.

• A 2-D multichannel image with size m-by-n-by-c, where c is the number of image channels. For example, c is 3 for RGB images, and 4 for four-channel images such as RGB images with an infrared channel.

• A stack of equally-sized 2-D images. In this case, A has size m-by-n-by-c-by-p, where p is the number of images in the stack.

Data Types: single | double | uint8 | uint16

net — Denoising deep neural network

dlnetwork object

Denoising deep neural network, specified as a dlnetwork (Deep Learning Toolbox) object. The network should be trained on images with the same number of color channels as A. The input size of the network does not need to match the size of A.

If the noisy image or stack of images A has only one channel and has Gaussian noise, then you can get a pretrained network by using the denoisingNetwork function. For more information about creating a denoising network, see Train and Apply Denoising Neural Networks.

Output Arguments

B — Denoised image

2-D image | stack of 2-D images

Denoised image, returned as a single 2-D image or a stack of 2-D images. B has the same size and data type as A.

Version History

Introduced in R2017b

R2024a: DAGNetwork and SeriesNetwork objects are not recommended

Starting in R2024a, DAGNetwork (Deep Learning Toolbox) and SeriesNetwork (Deep Learning Toolbox) objects are not recommended. Instead, specify the denoising network as a dlnetwork (Deep Learning Toolbox) object.

There are no plans to remove support for DAGNetwork and SeriesNetwork objects. However, dlnetwork objects have these advantages:

• dlnetwork objects support a wider range of network architectures which you can then easily train using the trainnet (Deep Learning Toolbox) function or import from external platforms.

• dlnetwork objects provide more flexibility. They have wider support with current and upcoming Deep Learning Toolbox functionality.

• dlnetwork objects provide a unified data type that supports network building, prediction, built-in training, compression, and custom training loops.

• dlnetwork training and prediction is typically faster than DAGNetwork and SeriesNetwork training and prediction.