Jump to content

Banner.jpg.b89429c566825f6ab32bcafbada449c9.jpg

Adaptive unsharp masking, some experiments


Recommended Posts

Given the lack of any new data to work on, I have written a simple addition to my Lucy-Richardson deconvolution code to perform adaptive unsharp masking, rather than the fixed form used by ImPPG. You really want to be able to tune the degree of sharpening to the S/N of the data. Indeed, I might even want to smooth the data at some locations in the image. There are various versions of this theme available in the digital image processing literature, but most that I could find either only sharpened using very small kernel Laplacians, or could only smooth adaptively (anisotropic diffusion). I decided to "roll my own" as a variant of the unsharp masking in ImPPG. The fixed form result on a VERY faint prom is shown partially inverted for clarity below.

Non adaptive unsharp masking, strength = 4.75, sigma = 1.15

file.php?id=14946&mode=view

As can be seen, the surface is fine, but the prom becomes very noisy (click to see full res).

I then made version which uses a couple of extra parameters: Max strength, min strength, cut-off grey level, and smoothing parameter for the "steering image". The idea is the following. We compute a steering image by smoothing the input image by some Gaussian. Wherever the steering image is low (well below the cut-off grey level), we assume the signal to noise if poor, so use a local unsharp masking strength close to the minimum desired strength. If the grey level in the steering image is high (well above the cut-off grey level), we sharpen with approximately the maximum strength. At the cut-off grey level, sharpening strength lies mid-way between the minimum and maximum strength.

Here are a couple of versions with max strength = 4.75, min strength = 0 (implying smoothing), sigma = 1.15 as before, and varying cut-off

Cut-off = 2200

file.php?id=14947&mode=view

Cut-off = 2450

file.php?id=14948&mode=view

Cut-off = 2700

file.php?id=14949&mode=view

Cut-off = 5000

file.php?id=14950&mode=view

Cut-off = 10000

file.php?id=14951&mode=view

Cut-off = 20000

file.php?id=14952&mode=view

As can be seen, we can progressively introduce more smoothing into the prom area, whilst retaining sharpening on the disk. The last two or three images are beginning to lose detail in the spicules, but the third in particular seems to get a good compromise between sharpening the disk well, without introducing noise in the prom. I am now using a fairly simple sigmoid function to steer between smoothing and sharpening, but other forms might be explored later.

Link to comment
Share on other sites

Looks good to me. As an image processing/analysis professional you'll want a routine that works globally on the image. Most amateurs will be happy to differentiate the needs of different image zones using layers and the eraser. (i can see you wince!!!)  However, I'm impressed and no mistake!

Olly

Link to comment
Share on other sites

Looks good to me. As an image processing/analysis professional you'll want a routine that works globally on the image. Most amateurs will be happy to differentiate the needs of different image zones using layers and the eraser. (i can see you wince!!!)  However, I'm impressed and no mistake!

Olly

Cheers Olly. it is true that amateurs might want to do things manually, but when processing 24 or more panes of a mosaic it becomes a right pain.

Link to comment
Share on other sites

Archived

This topic is now archived and is closed to further replies.

  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...

Important Information

We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue. By using this site, you agree to our Terms of Use.