Bringing out the details part 1: Local Histogram Equalization (Pixinsight)
The Local Histogram Equalization (LHE) process in Pixinsight is a great tool for bringing out the details in your image. It's a relatively simple tool to use. Pixinsight's website states that the purpose of the process is to "enhance local contrast and visibility of structures in low-contrast regions of the image".
I generally use LHE during the later stages of non-linear processing after noise reduction and a few curves adjustments. I like to use two iterations of LHE on my images:
to enhance the contrast of larger structures
to emphasize smaller structures
For this blog entry, I'll demonstrate the LHE technique I use on a nebula-only image as well as a galaxy image.
First up, the nebula image (NGC 2175 - the Monkey Head Nebula SHO palette). Here is a screen shot of my starting image that has been stretched, stars removed, noise-reduced, and adjusted with Curves Transformation:
The first thing I need to do is decide if a mask is needed. If there is a fair amount of background (dark regions) in your image, it's a good idea to mask those regions so that they are protected from LHE. You don't want to bring out any blotchiness in your shadow regions. I'll use the Range Selection process to create a mask that covers the nebulous region. In this case, I'll use the "lower limit" and "smoothness" sliders to get a mask that looks like this:
This should sufficiently protect my background.
Next, I'll apply this mask to the nebula image and then open the LHE process. The default settings should look like this:
I generally focus on two parameters when applying LHE: Kernel Radius and Amount. The Kernel Radius basically determines the size of the structures you'd like to enhance, while the Amount is simply how much of the effect you want (original/affected blend). I find that the "Contrast Limit" of 2.0 is good for most images. If you bump it up, it produces more contrast but at the expense of noise. I leave the "Histogram Resolution" and "Circular Kernel" at default, as this is what is recommended by Pixinsight - and I haven't found a benefit to changing these parameters.
LHE is very easy to overdo. If you preview the default settings, you'll get something like this:
Certainly not desirable - and you can see that over-applying this process will desaturate it and create plenty of noise.
I want to bring out the larger structures first, so I'll start by moving the "Kernel Radius" slider up until I see the larger structures that I want to enhance. I'm going to leave the "Amount" slider alone for now, so that the structures are easy to identify. In this case, I like the structures that appear around the Kernel Radius of 212 pixels. Here's what it looks like:
Now that I've determined the larger structure radius, I'm going to adjust the "Amount" down until I'm happy with the overall contrast level. I flip the preview on and off as I reduce the amount to compare to the original. In this case, I like the result at 0.190:
This is subtle, but it makes a difference. I'll apply this to the image, then proceed to target the smaller structures.
For smaller structures, I want to create a preview window that contains the structures that I want to emphasize. For this image, I've created it in this region that contains some smaller details:
Now, I'll reset my LHE process and use the Kernel Radius slider to determine the pixel radius that emphasizes the smaller structures I want to bring out. Usually this means 100 pixels or less (depending on your image). In this case, I find that a Kernel Radius of 42 pixels works well. Here it is, on full blast:
Just like with the larger structures, I'm going to reduce the "Amount" slider until I like the result. I'll also preview the entire image to make sure it doesn't negatively affect certain regions. I find that 0.160 works well for this image:
Once again a fairly subtle, but effective result. I'll go ahead and apply this to the entire image.
Here is the before and after LHE comparison:
You may find it necessary to add some saturation or curves adjustments after LHE to fine tune it. I bumped the saturation a bit to compensate for the contrast equalization.
Now for galaxies, my approach is very similar. Here's a luminance of M51, stretched and noise-reduced:
I'll go ahead and create a mask with the Range Selection process to protect the background:
Once I apply the mask, I'll create a closeup preview of the galaxy and target the larger structures. I find that a Kernel Radius of 140 pixels at the amount of 0.200 works well:
I'll apply this to the entire (masked) image and proceed with a second iteration of LHE to bring out the smaller details. Through experimenting, I find that a small Kernel Radius of 32 pixels at an amount of 0.180 looks good:
I'll apply this to the entire (masked) image. Here is the before and after LHE comparison:
Subtle, but if you look closely you can see many of the darker structures and fine details in the spiral arms are enhanced.
I hope you find this approach helpful. For the next blog on "bringing out the details", I'll show you how to find the structures inside the core of a galaxy using HDR Multiscale Transform. Stay tuned!