Noise Removal

From Wiki
Jump to navigation Jump to search
This article is out of date. You can help Panotools Wiki by expanding it.

Once the article is ok feel free to remove the {{Outdated}} template

There are plenty of programs available to remove noise from your images.

Here's a very incomplete list, if you know one you like and it is not listed, then please add it.

There are basically two ways to reduce noise in digital images: By filter in one image or by adding up several images.

By Filter

Reducing noise or grain in digital images by filter is somehow difficult, since the software has to automatically distinguish between image details and noise. The better products allow to create noise profiles in order to provide adjustments to that process.

A comparison with tests and a list of several tools of the Filter category is found on Michael Almond's page

Neat Image

Neat Image is shareware. It features both a standalone version and a photoshop plugin. It makes a profile of your camera that you can apply to a set of photo's. Neat Image's home page is

Noise Ninja

Picturecode's Noise Ninja is shareware. It filters out noise in 48 bit per pixel color space. It can batch process multiple images. It features both a standalone version and a photoshop plugin. Noise Ninja's home page is

Grain surgery

Grain surgery can add, remove or match grain (noise). It comes as a Photoshop plugin. Grain surgery's home page is


Noiseware is shareware. It comes as a Photoshop plugin. Grain surgery's home page is

Helicon Filter

Helicon Filters are shareware with a free edition limited in functionality. See for details.


Open source and very powerful filter collection. Use the anisotropic filter or any of the other ones to reduce noise. Available either as a command-line tool or a GIMP plugin.


Dcraw has a command-line option (-n noise_threshold) to erase noise using wavelets.

By Adding up several images

Adding up several images - either different scans from the same slide/negative or several digital shots from the same subject - are based on the fact that digital noise is random and different in each image but image detail should be the same.

With this technique the signal to noise (s/n) ratio increases by the square root of the number of images. This is you need 4 images to double the s/n ratio and 16 images to have s/n ratio 4 times as high as in one image.

Image Detail is not blurred, if the images are perfectly aligned.


A panotools helper application.


Averaging can be performed with Photoshop or any other image editor capable of layers by placing the images in layers and setting layer opacity to decreasing values (from top to bottom layer): 50% for top layer, 33% for second layer, 25% for third layer, 20% (=1/5), 17%(=1/6), 14%(=1/7), 13%(=1/8) and so on.