Neat Image is regarded as one of the better noise reduction programs though perhaps not in the same league as Topaz Labs DeNoise AI. It does, however, provide a free version having a limitation which I think can be acceptable for many uses in that it will only output a good, but not highest quality, 8-bit Jpeg. I normally produce some final versions of my images for use on the web or within our internal astronomy club competitions and these are Jpegs of around 1,200 to 1,600 pixels across. So if the noise reduction is done as the final step, having first reduced the overall size and bit depth, this is not really a problem.
It does imply a maximum image size of 1,600 x 1,600 pixels. However, in tests, it did appear to process the full area of an image 5,456 x 3,632 pixels in size. The filtered result was ‘differenced’ with the original and there was no difference in the result across the full frame.
It is best to use in ‘Advanced Mode’ rather than ‘Beginners Mode’ (click at the top right of the screen)
Having opened the image, one clicks on ‘Prepare Noise Profile” and, using the mouse select an area of the image that contains as little detail as possible. Then click on ‘Build Profile’ and Neat Image analyses the noise within the selected area. One then clicks on ‘Adjust and Preview’ and Neat Image applies what it thinks are the best parameters to remove the noise. These, and in particular, the ‘Noise’ slider can be adjusted if desired to adjust the level of noise removal.
There seems to be a possible problem when denoising astronomy images: how can the program differentiate between noise and stars? To investigate this, I have done a test using Neat Image to denoise an image of M42 (Image 1) to which I had added noise (image 2). The result is shown in image 3. I had used the noise reduction settings that the noise analysis had suggested. However, a better result was obtained by reducing the Noise slider from -15 to -35 (image 4).
I then used Starnett++ to separate out the stars and nebulae in the, noise added, image of M42 (images 5 and 6). I then used Neat Image to remove the noise from the Nebula image (image 7) alone to which I added back the stars to give me a second result (image 8). Obviously, the stars would not have affected the noise reduction applied to the Nebula image.
So I had 8 images:
1) The original image of M42.
2) The original image to which I had added noise.
3) The result of using Neat Image to denoise this image (so stars were included).
4) That of (3) but with the noise slider value reduced so giving a better result.
5) The Nebula image.
6) The Stars image
7) The denoised Nebula Image.
8) The result of adding the back the stars to the denoised Nebula image.
It is fairly obvious that the result (8) came closest to the original image suggesting that this might well be useful technique.
This program is worth a try – there is nothing to lose.