Noise Reduction — It really works A while back I started the search for noise reduction software. I polled the CCDNewAstro and CGEUncensored Yahoo! groups. The results of that search are on the wiki. There were four options people pointed to:

The first three on the list will work as plug-ins in Photoshop. Noise Ninja and NeatImage have means for procesing outside of Photoshop. They both cost $80 for the full version. Astro Actions has a bunch of useful items and it is only $20. Astro Actions does some noise reduction, but it is using standard Photoshop tools — it does not have a noise reduction process beyond the standard Photoshop filters. I did by it, however, as it has many good tools and I like supporting small software developers.

PixInsight, on the other hand, it an entirely different animal. A group of people are collaborating to bring a new astro-image processing tool to market. They have a fairly full-featured freeware version called PixInsight LE, and a time-limited core application. The time-limited application works for about a week at a time. This allows them to distribute fully functional software but still be able to sell a commercial version sometime in the future.

I have got to say that is is an amazing application. The learning curve is very steep, but you have access to capabilities that allow you to use base image enhancement algorithms with great control over the parameters. The noise reduction algorithm is quite good. And Lord knows, I need noise reduction since I image in such light polluted skies. The best analogy for the program is that it is like writing your own HTML rather than using a wysiwyg editor. Harder, but you have much more control.

Noise Reduction Example The image below shows the results of applying noise reduction to an image. I have cropped the image and reduced its size to limit download times.

When you roll over the image, you should see the pre-noise reduction version of the image. The change is remarkable. The graininess is gone but almost all of the detail of the image remains. The underlying image still needs some work, but used it as a training example for myself. In fact, as I look at the image, the color noise reduction may not be protecting the detail of the image sufficiently.

I do not really understand the math behind the noise reduction, but it looks at the image in small 3-pixel areas, and determines if it can smooth out differences between the pixels. It also looks for changes in brightness that would indicate the edge of an object or a star, and it protects those from being blurred.

Altogether, pretty cool. And so far, it is free (the commmercial version should be released soon).

Page updated: 3-Feb-2017 7:01:39 pm CST

bottom corner