Wierd NAFE program; image processor

[Kitsunegari]

So there is an interesting bit of software called NAFE that offer's alternative means to image processing.   It is supposed to be very good for hydrogen alpha images and was made for the SDO AIA assembly images.

 

Tried it out on one of my old calcium timelapse to see waht happens and it turned out a bit wonkey.  Surface looks like foam, which is not right.  But the limb looks "normal".  Overall it was interesting to see,  and affects on the gamma is a positive manner.

 

takes 

 

NAFE download link,   http://www.zam.fme.vutbr.cz/~druck/Nafe/System/0-info.htm

 

NAFE process.

 

 

 

IMPPG processing

[michael.h.f.wilkinson]

It does look a bit odd. I will have a look at the paper to see whether I can understand why

[GreatAttractor]

That's an interesting paper. If I understand correctly, NAFE is a sort of histogram stretch, performed piecewise for each image fragment.

19 hours ago, Kitsunegari said:

Tried it out on one of my old calcium timelapse to see waht happens and it turned out a bit wonkey.  Surface looks like foam

The example in the paper's Figure 2 looks similar to your result - i.e., dark, low contrast areas are "super-charged", stretched locally to cover a much wider brightness range. Certainly useful for a detailed analysis of small structures, but by definition it won't look like the raw image.

From the paper:

Quote

The program NAFE Image Analyzer, for MS Windows XP and later versions written in Borland Delphi, was developed to implement the NAFE method.¹ As the NAFE method is extremely time consuming, parallel implementation was necessary. Processing of a full (4, 096 × 4, 096 pixels) SDO AIA image takes approximately 5.5 minutes on a PC with Intel Core i7-3820 working on 3.6 GHz (8 processor kernels).

Sounds like a great candidate for porting to GPU!

Edited April 12, 2021 by GreatAttractor

[Peter Drew]

Possibly a typo, should read NAFF.  ?      🙂

[Kitsunegari]

4 hours ago, GreatAttractor said:

That's an interesting paper. If I understand correctly, NAFE is a sort of histogram stretch, performed piecewise for each image fragment.

The example in the paper's Figure 2 looks similar to your result - i.e., dark, low contrast areas are "super-charged", stretched locally to cover a much wider brightness range. Certainly useful for a detailed analysis of small structures, but by definition it won't look like the raw image.

From the paper:

Sounds like a great candidate for porting to GPU!

Most definitly;   Not really sure if it has any programable options to enahnce it furtherl   i tried tons of combinations in the .ini file but could not find anything useable outside of all zeros on the kerf just to enhance brightness.

 

Gpu processing would be great to increase the speed , but if the software is a lame duck for overall enhancement  without some slider buttons (similar to the wavelet sliders in registax)  i think it may not be worth the effort. It seems to function with very specific settings for the SDO images depending on wavelength.

 

3 hours ago, Peter Drew said:

Possibly a typo, should read NAFF.  ?      🙂

Well right now it has several different  typo names    The icon title saws NAWE.