HDRMT 'Issue'

[Warhen]

I'm sure someone has seen it occasionally- a spectacular compression and detail development in your preview, but a result which is not even close to it in the main image. In fact upon these occasions Median will result in no visible change.:<( This happens to me occasionally and when it does, there doesn't to be any remedy for it- neither Median nor default ATW will do it, nor will any combination of additional settings. It's frustrating because again, the preview is exactly what you want and as we know HDRMT is so darn good when it works. Any thoughts? Thanks!

[NGC7789]

How large is your preview relative to the whole image? And does it have sufficient representative structure? I have found if my preview is too small or the whole image has structures that are not represented in the preview then I may get very different results once applied to the whole. This is where assembling an image from a set of previews can be useful since you may not be able to get a good preview from a contiguous square preview.

[Warhen]

Thanks NGC, I'm aware that previews can act this way with certain processes. See attached. Previews 1, 2, and 3 give me a desired result. No perceptible change to the galaxy occurs in Preview 4 or the full image.

[Harry page]

Hi Warren

HDR does like to look at the full image to get a accurate preview , Also if there is no change I presume you have gone non linear

I,e it will have little effect unstretched

Harry

[Warhen]

Oh yes of course Harry, it's a nonlinear LRGB. I see this every once in a while- brilliant result in the preview, nothing close to it on the actual image, or the larger preview for that matter. See attached guys with same settings. Preview 1 is smaller than Preview 2. I cropped the galaxy out of a clone and HDRMT does work on the crop. So, it seems to be a size issue (?) and not disparity between the preview and the image.

Well, in the end, I applied HDRMT to a masked crop of the galaxy, then used DynamicAlignment to register it back to the main image, then added the result through a mask with PixelMath. That was a lot of work. I would LOVE to know why this happens and if there is any way around it. Thanks!

[rbotero]

I reported this under Bugs forum but Larry pointed me to this post.   I am "glad" the Masters (Harry, Warhen) are looking into this!

For my part, since I only needed the Ha information to complement my RGB data, I used the SubstituteWithPreview Script to add my separately processed Previews to a new Target image.

Roberto

[Warhen]

Masters, schmasters Roberto- you are a pro yourself! I forgot about this script and it is a great workaround here! Would still like to know the cause of the HDRMT problem though to avoid the extra steps. For that we need Juan!

[NGC7789]

Note that Juan has replied in the bug variation of this thread

http://pixinsight.com/forum/index.php?topic=8461.msg55085#msg55085

[Carlos Milovic]

It is a feature, not a bug.
Jokes apart, this is normal in such multiscale operations. You should try with a different number of layers and see if you can replicate the results with a smaller preview. Also could help to separate sources (DSO and stars), and HDR process then individualy.

[Warhen]

Thanks Carlos. Mine is sort of opposit- good results in small preview, nothing in parent image. Thanks for the SubstituteWithPreview script! Thanks NGC, read that and replied. Mind of I ask who you are NGC?