View Full Version : Processing raw images exercise for October 2012

Roger Clark
10-13-2012, 08:25 AM
This month I am contributing one of my images: a night sky image around the constellation Cygnus. It is a 10 second exposure with a 24 mm f/1.4 lens at f/2, ISO 1600. The goal is to bring out the faint detail detail and maintain as much star colors as possible. Note too some hints of light pollution in one corner (faint orange glow). In particular, bring out the detail and color in the North America nebula, which you can see near center.

So show us your best on what you can do with this image. As with all images, BPN guidelines apply. You will need to process the image and create a jpeg not more than 1024 pixel wide to show your work. The photographer retains all rights, so this exercise is only for posting on BPN unless you get written permission from the photographer for other uses. All posted examples need an explanation of what you did. The more detail you can give, the better, as it can help others to better understand better your steps. Examples are in the other raw processing threads (to which you may still contribute). Also consider posting a 100% view of a small section to show the details.

The raw image is located here: http://www.clarkvision.com/bpnraw/roger.clark.night.scene.C45I1088.CR2

You may also process and post results from previous months. Last month's image did not get a big response. Was it too hard?


Roger Clark
10-13-2012, 10:01 PM
All photographers should check this out. Even if you are not into night photography, this month's exercise will show how to dig signals out if the faintest parts of an image. Thus, it has applications to getting shadow detail out of any image, whether the black feathers on a bird in shadow, detail in a landscape image shadow, or the night sky. The exercise also has implications for preserving colors near the saturation limits of the camera, so again has applications to all types of images. Download the raw image (link above) and give it a try.


Andre van As
10-16-2012, 06:23 AM
Hi Roger
Although I do not have a clue what the real magnified night sky should look like I had a go at your exercise to learn more about PP
This is what I did in ACR (v6.1)
Tint +9 - gave a nice blue appearance - but I sacrificed reds
Blacks +9 - this brought out more stars in the BG
Brightness +150 - perhaps too much but more became visible
Sharpening +64 - this produce more background detail
Radius 2 = ? too little
Detail 25

NR no adjustments as stars disappeared when I input values - seems that stars were interpreted as noise




John Chardine
10-16-2012, 03:42 PM
Yes it was too hard!!!! I'm going to have a go at this one!

Tom Graham
10-16-2012, 08:04 PM
It's too hard for me!! I don't know what "reality" is for this. I went outside last night and did not see anything celestial but the moon and venus (here in beautiful sunny southern California).
I did play with exerecise anyway and decided that there is likely a faint "orangenus" in Cygnus. But I'm not going to post it because of the obvious black halo around bigger stars. Hope you, Roger, will show us yours.
Regards - Tom

John Chardine
10-16-2012, 08:59 PM
Here's my go at it. Very interesting exercise, not really knowing what the target should look like. I did the major processing in ACR as follows:

Exposure +1
Clarity +50
Vibrance +50
Blacks -40
Whites +40
Contrast +15

Then resampled and sharpened with Smart Sharpen (Lens blur) 80%, 0.2.

My general tactic was to bring out the detail and the colour that was in the image to begin with.

Roger Clark
10-16-2012, 09:52 PM
Here is a quick attempt using gimp in linux. What I did was use the out of camera jpeg (which is also in the directory link above, just replace CR2 with JPG). I used curves to increase the mid tones, set the black point in red and green slightly higher, then unsharp mask with radius 34, threshold 5, amount 0.75 (I think this would be 75 in photoshop). Finally, boost saturation about 12%.


Roger Clark
10-17-2012, 10:34 AM
I'm attaching histograms for your image. First is a linear histogram (like that in photoshop) then log histogram. I generated both of these with gimp.
After I've posted histograms for each image, I'll discuss results.

Roger Clark
10-17-2012, 10:36 AM
I'm attaching histograms for your image. First is a linear histogram (like that in photoshop) then log histogram. I generated both of these with gimp.
After I've posted histograms for each image, I'll discuss results.

Roger Clark
10-17-2012, 10:39 AM
I'm attaching histograms for my posted image. First is a linear histogram (like that in photoshop) then log histogram. I generated both of these with gimp.
I'll discuss results in another thread.

Roger Clark
10-17-2012, 11:01 AM
Now let's examine the histograms for the 3 images. Note the blue channel on the left side (the low end). All 3 images (Andre, John, and mine) show some of the blue channel is clipped, and the amount clipped is in the same order as posted. Same with the green channel. For the red channel, mine is not clipped on the left side, but the other two remain clipped. I adjusted the black points for the red and blue channels and overshot a small amount for the green (I shouldn't have). I did not adjust the black point for the blue channel--that is a property of the jpeg. I should process the raw data to avoid that. At the high end, all images show saturation in all channels, and this is due to the huge dynamic range stars have in a typical scene and all the bright stars get saturated. So to avoid further saturation, one should not use the right side of the levels tool, nor increase exposure in raw conversion.

I had a conversation with another photographer recently who does nightscapes, and he chooses to process his images to give the stars a blue color, blue Milky Way and blue sky background as he does not like the brown color of the Milky Way. While that is a choice, it is not close to real color. The Milky is reddish-brown due to interstellar dust. There also emission nebulae (like neon signs), and they appear red (a few smaller ones appear green or blue). For example, the North America nebula near the center of the image should appear reddish west of the Mississippi and blue-ish east of the Mississippi. The reasin is that west of the Mississippi is hydrogen-alpha emission from the nebula, and to the east is star light reflecting of of dust by Rayleigh scattering make a blue color (like that for our blue sky).

So in processing, it is important to not clip the blacks in any channel, nor highlights, if one wants to maintain some detail in the highlights and lows. So in the raw converter, I set the black point to zero (e.g in ACR) and work on the lows in the image editor where I can use curves with more control, and I always monitor the histogram. Often in conversion from Adobe RGB to sRGB some highs and lows get clipped. That is an unfortunate side effect and one has to determine if that harms the visual image quality.

For images like this, it is best to have the histogram bell-shapes for all three channels separated from the left edge (this is true for most images), but for star images, the histogram can be scrunched up close to the left edge as long there is no significant clipping.