PixInsight Forum (historical)

PixInsight => Bug Reports => Topic started by: Nocturnal on 2009 March 10 20:10:18

Title: HDRWaveletTransform: meaningless trials on large previews?
Post by: Nocturnal on 2009 March 10 20:10:18

Hi,

I have an image of about 3K by 2K of mostly background but with a few galaxies sprinkled on. I'd like to use HDRWT to enhance contrast/detail in the bright cores of the galaxies. I defined a preview of 1000 x 900 pixels around one galaxy and applied a 2 layer HDRWT. I liked the result and applied it to the main image. The galaxy is hardly touched. Only the background was changed and not for the better (it was smooth, now it's mottled).

I understand (somewhat) that wavelets look at structure but the first two layers look at small scale structure, right? If so then why does a 1000x900 section behave so differently then the 3000x2000 image?

I realize this is probably not a bug but I'm not sure where else to post this.

Small notes on HDRWT:

- there are no tooltips for the two sliders in the Deringing section. One for 'isolation' would be nice :)
- there is no description for the process in the process explorer

Title: HDRWaveletTransform: meaningless trials on large previews?
Post by: Juan Conejero on 2009 April 10 09:33:19

Hi Sander,

Indeed the HDRWT algorithm is not the best friend of previews. I'd need to play with the image in question to give you a useful answer.

Quote

there are no tooltips for the two sliders in the Deringing section. One for 'isolation' would be nice

Of course I agree. I'll add those tool tips for 1.5.

Quote

there is no description for the process in the process explorer

...as happens with most processes. Nevermind; the Process Explorer window has been refactored for 1.5 and it now doesn't show that kind of descriptive documentation. It now provides a property tree for each process, with information mainly oriented to developers.

Title: HDRWaveletTransform: meaningless trials on large previews?
Post by: Nocturnal on 2009 April 10 09:42:57

Hi Juan,

I thought the description for some of the processes was quite informative so I look forward to seeing how you've re-implemented this. Thanks for adding the tooltips.

Title: HDRWaveletTransform: meaningless trials on large previews?
Post by: Harry page on 2009 April 10 10:31:57

Hi Sander

I have found if you use a simple luminance mask on the complete image the preview seems to work much better and if you clip the mask a bit the faint edges are protected and you don't get that grainy effect
on low signal areas
Worth a try anyway


Regards Harry

Title: HDRWaveletTransform: meaningless trials on large previews?
Post by: Nocturnal on 2009 April 10 10:35:42

Hi Harry,

shouldn't that be the same as checking the 'luminance mask' option in the HDRWT process? I will give it a try though. Thanks for the tip!

Title: HDRWaveletTransform: meaningless trials on large previews?
Post by: Harry page on 2009 April 10 11:02:20

Hi sander

I know what you are saying but i don't know if by clipping the mask you get better protection!!!!

Regards Harry

Title: HDRWaveletTransform: meaningless trials on large previews?
Post by: Juan Conejero on 2009 April 10 11:35:32

Hi Harry and Sander,

HDRWT's integrated luminance mask isn't exactly the same as a common (image) mask (because the integrated mask works with temporary pixel values in a different range). OTOH, you gain more control with an image mask because you can modify it with histograms, etc. Finally, there's no problem at all in using both masks simultaneously. HDRWT will do its job and then the image mask will be applied to the final image.