Noise Reduction in Low Signal Areas Only

[Juan Conejero]

can you elaborate on how to go about building a linear mask for linear noise reduction using PixelMath.

As implemented in the MLT and MMT tools, a linear mask is just a duplicate of the linear image multiplied by an amplification factor, so the corresponding PixelMath expression can be as simple as:

$T*k

with k defined as a constant symbol such as k=100 for example. In the case of a color image one can apply this either to the duplicate RGB image (because in PixInsight you can use RGB masks with RGB images), or to single components such as H or I (extracted with ChannelExtraction working in the HSV and HSI color spaces, respectively). The linear mask should always be low-pass filtered to soften edges and make it more robust to local variations; this can be done easily with the Convolution tool applying a Gaussian filter.

The MultiscaleLinearTransform and MultiscaleMedianTransform tools include a linear mask generation feature that can be used to automate this process. If you enable the Preview mask option (Linear Mask section), you'll get a linear mask directly if you apply the process to a duplicate of the image. This is similar to the same feature that has been available on the ACDNR tool for a long time. The main advantages of this method are that you can use the Real-Time Preview functionality of the tools, and a good linear mask is much easier to build because the tool computes it adaptively from image statistics.

can this linear mask be exported in order to use it in other processes?

Definitely yes. The linear mask generated with the MLT or MMT tools is just a regular mask. It can be used with any tool, and not just with noise reduction tools. The purpose of a linear mask is to map the SNR of a linear image, and hence it can be useful for any process that should be applied proportionally (directly or inversely) to SNR. For example, a noninverted linear mask can be useful for deconvolution.

We'll probably release a LinearMask tool in September to simplify generation of these masks.

[Juan Conejero]

Of course, listening to Juan and the debate, clearly there is a bit of a religious schism at work here too.

Please don't see my opinions as religious statements. They are logical productions of the way I understand image processing and astrophotography. As I understand it, not everything is valid in astrophotography. Hand-painted masks and hand-driven procedures belong to the category of unacceptable practices (an exception is correction of small cosmetic defects with the conditions I described in my previous post). The why and how are more important than the final product in astrophotography, but this involves reflection and understanding at each stage of the whole process. Unfortunately, there are plenty of hand-painted pictures in the astrophotography community these days, including a significant fraction of APODized images.

[mads0100]

Juan,

     Thanks for the detailed explanation.  Helpful as always!

Chris

[jkmorse]

Juan,

May have made a poor analogy, since I agree my comments in no way involved religion, just reflecting on the difference in philosophy that you document in your response so well.  What I see as the true glory of PixInsight is that it drives you to understand and appreciate your data in a way that "painting pretty pictures" never can. 

And thanks for explaining the issue of PixelMath masks so well.  As always, you make the complicated understandable.

Best,

Jim