Hi all,
Just a quick knock up of some data from Astrofest. I finally got some colour subs after a marathon N.B. effort of late. This is NGC6559 and associated IC objects.
Details are: LRGB 100, 90, 90, 100, TSA120, SXVR H-16, Maxim 5, PS 4. Any constructive criticism or processing tips would be most appreciated. See below for a direct link to the image.
Thanks for looking. Darrell.
Impressive work Darrell. The framing of the scene grabbed me at first, very nice. The data looks black clipped however as the shadows are completely black. Perhaps you were looking for more contrast. Keep an eye on this histogram as you process. I think you could improve on this data set.
Very nice Darrel. I do have to agree with Jase but don't think it detracts from the image very much at all.
The blue in the nebula is very subtle and smooth.
One for the pool room. Would look great framed and up on the wall.
Well done.
Thank you all for your kind and fair coments regarding this image. Jase, good call, I should keep the histogram window open while processing, keep an eye on the black point My focus was on sky background levels which seem way brighter than the very thick dust. It seems to be a real balancing act Anyway, thanks again, cheers and clear skies!
That's a beauty Darrell. Nice and sharp and detailed. Its not that bright an object and you display it very clearly. The background though seems to have a slight excess of green in it throwing the colour a tad.
There seems to be quite a few people on these forums who erroneously claim some people's images a clipped without actually inspecting the images. More often than not, they are incorrect and just viewing these images on an improperly calibrated monitor (laptop screens being the usual suspects).
Please don't make these erroneous claims guys - it is very unhelpful, especially for beginners who though they were doing it right and are now confused (again).
Detecting a clipped image is quite well defined. A clipped image will have R G B values that are exactly 0, 0, 0 (even then, that value is allowed if it represents the absolute blackest pixel in your data set). I cannot see these in this instance. You can argue about the preferred background levels, but this has nothing to do with clipping.
Ivo, assuming you are referring to my statement below;
Quote:
Originally Posted by jase
...The data looks black clipped however as the shadows are completely black...
More than happy to take this offline, but since you've gone public I'll respond accordingly.
Quote:
Originally Posted by irwjager
More often than not, they are incorrect and just viewing these images on an improperly calibrated monitor (laptop screens being the usual suspects).
Completely agree. I often laugh at those that process images on laptops. Such screens don't calibrate well either. They are better off plugging in an external monitor. For the record, each image I comment on in this forum I load into photoshop. Sounds ridiculous I know as its rather time consuming, but it is the only way to truly appreciate ones work, against an accurately calibrated monitor profile. I don't trust web browsers.
...alas lets cut to the point;
Quote:
Originally Posted by irwjager
A clipped image will have R G B values that are exactly 0, 0, 0 (even then, that value is allowed if it represents the absolute blackest pixel in your data set). I cannot see these in this instance.
Not correct. You are making the assumption that RGB is uniform when its not. An individual channel can be clipped. This is histogram 101, am sure you realise this. Suggest you check the green and blue channels. More than happy to be corrected - surprise me! http://www.iceinspace.com.au/forum/....es/happy19.gif
....
Darrell, I wish to apologise for not being specific about the green and blue channel clipping in the original post. If anything this is the misleading information and are not 'erroneous claims'. My intent is to drive improvement through constructive feedback, no different to what I've done it the past for others. Some take it, some leave it.
There seems to be quite a few people on these forums who erroneously claim some people's images a clipped without actually inspecting the images. More often than not, they are incorrect and just viewing these images on an improperly calibrated monitor (laptop screens being the usual suspects).
Please don't make these erroneous claims guys - it is very unhelpful, especially for beginners who though they were doing it right and are now confused (again).
Detecting a clipped image is quite well defined. A clipped image will have R G B values that are exactly 0, 0, 0 (even then, that value is allowed if it represents the absolute blackest pixel in your data set). I cannot see these in this instance. You can argue about the preferred background levels, but this has nothing to do with clipping.
Cheers,
Yeah Ivo in this instance it is black clipped in the blue and green, even a little in the red.
eek! Didn't mean to cause a stir!
I certainly wasn't targeting anyone in particular and I apologise if I came across dismissive of anyone. I'm was just trying to look out for the newbies!
I don't trust histograms because of this; big black stacking artifacts (as is the case here at the bottom of the image) skewing your histogram. Just like you Jase, I always load the image into a processing application. When I suspect clipping, I always sample the darker areas for R G B values with a dropper. Doing this the image looked ok to me at the time and nowhere could I see perfectly black pixels or distinct colored tinges.
Looking at the histograms, cropping the stacking artifacts out, things look a bit better. Though, I have to admit, still not perfect for the green and blue channels.
However, to illustrate the distribution of (possibly!) clipping pixels in the green and blue channels (assuming we are now satisfied red is not a problem), I've created an image showing where pixels are (possibly) clipping the green and blue channels (e.g. they are both 0, with only red being positive - done by removing the red channel, then doing a threshold of all but the 1st value in the histogram).
Not too shocking now I would say. Seeing as their distribution is quite sparse (e.g. there are non-clipping pixels adjacent to most of the 'clipping' pixels. I would say that a good deal of these pixels are 0 because they are just properly 0 or fractional - not negative. Indeed, if I use a morphological dilate operation, virtually none remain (e.g. proof that they are very local 'dips' in the signal and bounce back up to connect with a non-0 R, non-0 G, non-0 B neighbouring pixel). Indeed in a 16-bit resolution these pixels could well be non-0.
In this synthesized image, I would be worried if there were a clear boundary beyond which large black areas were visible, however there are always non-black pixels peeking through here.
To make a long story short, I still believe the full dynamic range was used in this image, even though there are pixels that have either red, green or blue set to 0. I don't believe it is to the detriment of the image and I don't believe any detail is lost. Now, would I have liked to have seen some background level? Probably. But I still don't think there is a case to call the image 'clipped'.
Again, my sincere apologies if I offended anyone - happy to take any further conversation private.
It's all good Ivo. If you go in PS and apply an adjustment layer as a curve then select the black color picker and press ALT while hovering across the picture you'll see the clipped pixels highlighted. You can do that for white clipping as well and per channel R, G or B. But you are also right in saying that an histogram can represent outliers as well and the clipping is not always valid data.
It's all good Ivo. If you go in PS and apply an adjustment layer as a curve then select the black color picker and press ALT while hovering across the picture you'll see the clipped pixels highlighted. You can do that for white clipping as well and per channel R, G or B. But you are also right in saying that an histogram can represent outliers as well and the clipping is not always valid data.
I guess it's typical of a low-level coder; we start counting at index 0, we don't regard 0 as the 'off-limits clipping flag'.
My focus was on sky background levels which seem way brighter than the very thick dust. It seems to be a real balancing act
Anyway, thanks again, cheers and clear skies!
Didn't mean to cause a stir!
If you go in PS and apply an adjustment layer as a curve then select the black color picker and press ALT while hovering across the picture you'll see the clipped pixels highlighted. You can do that for white clipping as well and per channel R, G or B. But you are also right in saying that an histogram can represent outliers as well and the clipping is not always valid data.
