I was reading a thread this morning and realized I had no idea what was going on. Can someone explain what "Clipping" is and what it does to an image? Thank You
RE: Clipping
I was reading a thread this morning and realized I had no idea what was going on. Can someone explain what "Clipping" is and what it does to an image? Thank You
Hey Paul, there are a couple of types of "clipping"
One is where you have overexposed highlights rendering those "blown" or "hot" and usually detail-less. If you have your blinkies turn on in your camera you will see these flashing when reviewing your images on the LCD. On the histogram you will have data so far right of the right that those are touching, or "clipping", the edge of the histogram. This happens with colours too such as yellows and reds so watch the RGB histogram.
The other way of this happening is during post-processing when someone overzealously oversaturates the colours resulting in clipping of some colours (again usually reds and yellows), or overdoes the contrast...when looking at the resulting histogram, the same thing happens at the far right edge of the box.
Another easy way to check is in PS Levels...press and hold down the "ALT" key, and click on the highlights slider (the one on the right, directly beneath the histogram). If anything other than solid black shows up, then you have clipping going on...
We could write a whole chapter on this phenomenon, but this is about it in a nutshell:)
Last edited by Daniel Cadieux; 10-17-2010 at 10:24 AM.
Thanks Dan...So if I get this right, when processing and image; either colors or exposure, its a way of knowing the proper amount of adjustments to give in ACR, by eliminating or keeping the clipping to a minimum? :confused:
Thanks Daniel.
In ACR, if you look in the upper left and right hand corners of the histogram, you will see 2 triangles. These represent the clipping of shadows and hightlights and will change colors when you have clipped something. Clicking the triangles will show you the pits that are clipped on the image.
Originally Posted by Paul Lagasi
Chris
0 .· ` ' / ·. 100
I have a high sarcasm rate. Deal with it.
include('sarcasm.php')
Paul: Another way to describe clipping is this: In a digital image you have three channels: Red, Green and Blue and each is assigned an integer value between 0 (black) and 255 (pure Red, Green, or Blue, respectively) (0-255 for an 8-bit image, where 2 raised to the 8th power gives you 256 unique values per channel -- that is, 8 bits). A 16 bit image is basically the same, except there are 2^16 unique values per channel (0 to 2^16-1). The Hue, saturation, lightness and LAB color models are just mathematical transforms of the RGB model. The carry-away point is that there is a finite number of values available for any given channel. If you overexpose, underexpose, or otherwise shift the histogram too far in either direction, darker areas will get set uniformly to 0 and lighter areas set to 255. Once that happens, information is lost -- different pixels that may have ranged from, say 245 to 255 are all smushed to the single max value of 255 -- from which there is no way to separate them back to the original 245 to 255. The 10-values between 245 and 255 are all now assigned 255 and there is no way to determine what those affected pixels started out as. They have been clipped.
Paul,
It is the image that gets judged not the histogram. Some times it is necessary to clip R/B/G channels to reproduce exact representation of the natural scene.
-Sid
I would very much like to see an example of this Sid.
John,
I can show you an image of the year on BPN.
Not sure when the clipping occurred during captured or conversion to sRGB but this image shows clipping in the red channel as posted
http://www.birdphotographers.net/for...-Francis-Bosse
and this one
http://www.birdphotographers.net/for...9-Kobus-Tollig
also this one
http://www.birdphotographers.net/for...a-van-der-Walt
some images even have clipping at both ends of the histogram
http://www.nanpa.org/photos/showcase/2011/00432446.jpg
Last edited by Don Lacy; 10-17-2010 at 06:39 PM.
Don Lacy
You don't take a photograph, you make it - Ansel Adams
There are no rules for good photographs, there are only good photographs - Ansel Adams
http://www.witnessnature.net/
https://500px.com/lacy
Sorry to be pedantic guys but what I was asking Sid for was an example of an image where "it is necessary to clip R/B/G channels to reproduce (an) exact representation of the natural scene". I wasn't questioning whether an excellent image like the one above could be made if clipping was present.
Don- I don't doubt there are lots of images posted out there that show clipping.
OK, the reason I asked for an example must be obvious- I don't agree with Sid's statement. Our eyes have a much wider dynamic range than digital sensors of today so how can a digital image that shows clipping in either the R, G, or B channel, or in the highlights or shadows, reproduce an "exact representation of the natural scene"? And I am assuming that "exact representation" means what the person observed rather than what I could measure with a spectrophotometer. Do our eyes clip in the colour channels? I can think of one example where our eyes may clip the shadows- in a brightly lit scene with lots of highlights- say a snow field-, I can imagine our pupils shutting down so much that no detail is visible in shadow areas. However, whether you would ever want to make an image of such a scene is another matter.
Surely if you took two identical images, with the exception that one showed clipping and the other didn't, the latter would be preferable? Given all this it must be preferable to reduce to a minimum or eliminate clipping altogether.
I'm questioning to learn here, not to be obtuse or combative.
Last edited by John Chardine; 10-17-2010 at 08:36 PM. Reason: spelling
Hi John I think we are looking at this from two different angles I agree you do not want any clipping at the time of capture and I do not want clipping in my working copies of a file in the adobe RGB color space. I will accept clipping when converting the image to sRGB for web presentation as long as there is no color shift and the reason for the images I chose to link. I read Sids comment to mirror my own thought of converting images for web presentation.
Don Lacy
You don't take a photograph, you make it - Ansel Adams
There are no rules for good photographs, there are only good photographs - Ansel Adams
http://www.witnessnature.net/
https://500px.com/lacy
Fair enough. We all believe in what we do
Yes, I agree with you on this.
Then let me ask you this. What would happen to the image info regarding out of digital spectrum but capable of human visual spectrum? Do you want me to completely throw it off since histogram is not capable of showing it?
What I wanted to say about Exact representation is since photographer experienced the colors he should try to reproduce it even though they are out of digital camera spectrum.
I got to disagree with you on this too. In bright light snow and white sand works as reflectors and eliminate shadows. This image was shot at 4.30PM with a bright sun. Sand worked as reflector and eliminated the shadows.
http://www.birdphotographers.net/for...010-Sid-Garige
Some times. If photographer can reproduce color tones closer to original color tones.
We all got our style we should stick to it as long as if we think it is right.
You dont need to say that. I know you are little more logical than me. :)
Last edited by Sid Garige; 10-17-2010 at 08:42 PM.
Wow..I had to read, this thread a few times....Thank You, all very much..
I'll add several things here.
People say reds and yellows often get clipped. So does the blue channel with skies. Sometimes even though one channel is clipped, the tonality in the other channels (if the color is not too skewed to one color channel) will show enough variation that we still perceive detail.
I've done some experiments on the dynamic range of the eye (I need to write it up on my web site). I measured the dynamic range in viewing any one scene to be 14 stops, consistent among several people. DSLRs have close to 12 stops per pixel, and in a small region, we can see about 14 stops in a processed image. With highlight recovery, it can be pushed a little more. HDR images can, of course, far exceed that range. But DSLRs are generally close to the dynamic range of the eye.
Roger
OK Roger, thats's very surprising. I wonder why it is that scenes to my eye almost always look less contrasty than they end up out of the camera? Is my brain filling in some detail perhaps? Right now I'm looking out my kitchen window to a brightly-lit, light-toned shed across the garden and some spruce trees in shadow in the foreground. I can see tons of detail at both ends but my 1D4 at ISO 400 cannot do it. I exposed so that the shed was not clipped and the image is very blocked in the shadows with the histogram piling up on the left hand side. Because I am inside, my pupils are probably quite open so that explains why I can see detail in the shadows, but how come the shed doesn't look bright white and clipped? I know this is n=1 and you have to trust me describing what I can see, but there it is.
I found this paper on the subject:
http://portal.acm.org/citation.cfm?id=1836248.1836251
and the abstract mentions that under the conditions of the experiment human vision has a dynamic range of 3.7 log units, which I assume means 3.7 orders of magnitude. This seems to be quite a bit less than 14 stops but I have to admit to being a little unclear about what 3.7 log units means.
From the foregoing discussion and prior experience, I think I have some understanding of what clipping is and how it can occur. But my question is this: If clipping is detected in a histogram following conversion of an image file to sRGB for posting, and if no color(s) are blown out, what must be done? Is it necessary to try to correct the sRGB file, so that the problem in the histogram is eliminated? Or are there some circumstances in which it can be ignored.
I'd really appreciate practical guidance on this issue, because I'm still not certain how to handle it. :confused:
Norm
Human visual system
Its fitting that HDR imaging is being brought up in this discussion, because texts on the subject go into great detail on the human visual system and HDR tone mapping. See, for example, the excellent text here -- it is not an example of practical guidance on the issue, but very good if your bent is science, engineering and computer graphics.
Paul: What hath you wrought!:D
Last edited by David Gancarz; 10-18-2010 at 07:54 AM.
John,
The 3.7 is 10^3.7 = 5012 and that is 12.3 stops, which seems kind of low. But it does depend on how far apart different levels are. In my tests, I used a window with as series of gray to black patches indoors comparing to the outside scene, then did spot metering of different spots. As bright/dark subjects appear closer together, dynamic range will drop due to scattering in the eye. This also happens with optical systems, and some published dynamic range tests of digital cameras suffer from scattered light, lowering the result. I see this a lot with imatest results.
Here is one test that illustrates the dynamic range (and exposure latitude) of DSLR images:
http://www.clarkvision.com/articles/...e-1/index.html
Note the +2 stops images down to the -8 image (which includes detection down to -14 stops). One can pick the noise level that is acceptable, and see what dynamic range works for you. Of course, that is with a 1D2. I should redo that test with a 1D4.
Roger
Norm,
We had a big discussion on this recently, I think in the macro forum if I remember correctly. OK, here:
http://www.birdphotographers.net/for...read.php?70996
and guess what, it was started by you. :D So did we not cover the subject well enough? My take on it was that clipping happens in conversion to a smaller color space. It might be better to fix the problem, but sometimes the impact on the image is too much, so it is better to let them clip. As with many things in photography, it comes down to personal preference.
Roger
Agree with Roger completely. As an image maker you are already aware of the color tones of the subject & background. Also with experience you will learn to not over saturate the image and assuming you work with a good calibrated screen. With this you are in a position to judge the color tones yourself and not limited by histogram. If it clips, let it clip.
Last edited by Sid Garige; 10-18-2010 at 09:51 AM.
Very timely thread as I was standing in the field yesterday trying to decide the best avenue of approach. So how would you handle the following scenario?
What do you do when the histogram shows that one channel is clipped (like blue or red) but the combined histogram looks pretty good? Do you let the channel clip and lose the detail there or do you reduce exposure to remove the clipping, which means that the luminosity exposure shows underexposure? I realize I can always bracket but was wondering which will ultimately produce the best file. Yes, shooting in Raw.
Sandra,
With out seeing your image and channel histograms its almost impossible to tell you what to do.
Best thing would be to make the following images and compare them side-by-side.
1. Good combined histogram with one clipped channel.
2. Reduced exposure to avoid clipping
3. Corrected channel clipping.
make all 3 images and place them side-by-side. That will tell you the best way of dealing with that particular image.
-Sid
Thanks, Sid. No general "rule of thumb?"
Roger:
I'm OK with that. I raised the issue here only because one individual in that macro forum discussion implied that posting an image exhibiting any histogram clipping might not comply with the guidelines for posting on BPN. As long as it's OK to use reasonable discretion in such cases, and posting of an image that looks good to the poster but may have an imperfect histogram (in the context of this discussion) is not a violation of some guideline, that's fine with me. ;)
Norm
My personal preference is to toss the image or avoid the conditions that caused the clipping to begin with.
Sometimes you just have to let go.
Good one Norm! I'm not aware of that guideline myself!
This is a fascinating discussion though with a crucial turning point being Sid's admonition to basically ignore the histogram and go with the look of the image as the final arbiter. The thread seems to have now evolved to the point where we should not care about clipping at all anymore (a little hyperbolic I agree!).
With all the talk and emphasis on avoiding clipping in myriad past BPN posts, you can understand that I am getting confused. And if I used smiley faces I'd put in a confused one here, but I don't so I won't.
Last edited by John Chardine; 10-18-2010 at 11:06 AM.
Norm I tried to be diplomatic during that discussion but that advice is pure nonsense as far as I am concern and the fact that at least three images chosen for image of the year honors show clipping in the sRGB color space as posted should be evidence enough that it is irrelevant in most cases. Sometimes there are color shifts and I will correct them and the reason I convert to the sRGB color space before using the save for the web option in PS.
Don Lacy
You don't take a photograph, you make it - Ansel Adams
There are no rules for good photographs, there are only good photographs - Ansel Adams
http://www.witnessnature.net/
https://500px.com/lacy
There can be time when you want part of the image to be pure black or white, therefore you would have clipping in the histogram, does it really matter?
With all the talk and emphasis on avoiding clipping in myriad past BPN posts, you can understand that I am getting confused. And if I used smiley faces I'd put in a confused one here, but I don't so I won't.John, I try to avoid clipping at the time of capture at all cost every once in a while doing so negatively impacts the rest of the image in those cases I will try to correct the clipping in post. The following image is an example were I had clipping on the breast and head but correcting it in the field would have underexposed the rest of the image losing detail in the wings. In ACR I ended up having to process the image twice once for the total image and a second time to pull detail at of the clipped areas I then combine both versions in PS using a layer mask. When converting the image to sRGB for web posting the histogram showed the red channel clipping again but since their was no color shift I ignored it. Now if I would have let this go in the field or given up on it in post I would not have a Scarlet Tanager image in my files. The limitation of capturing light on a sensor or a piece of film forces you to make compromises understanding those compromises and the effects they have on your image lets you make informed decisions and on occasion that decision might be to let the highlights or shadows go. When it comes to converting an image to sRGB for web presentation so long as the image looks the way I wanted it to in adobe RGB I ignore the histogram and any clipping that it shows. Speaking for myself only I would rather have an image that has an emotional impact on the viewer even if it might not be technically perfect and we should not let the process get in the way of the vision if that makes sense.My personal preference is to toss the image or avoid the conditions that caused the clipping to begin with.
Sometimes you just have to let go.
Don Lacy
You don't take a photograph, you make it - Ansel Adams
There are no rules for good photographs, there are only good photographs - Ansel Adams
http://www.witnessnature.net/
https://500px.com/lacy
We cannot avoid clipping 100% of the time, but if it can be corrected without affecting the image then that is the best option IMO. Some minor clipping, if unavoidable, is OK to the eye...but when it is too much it negatively affects an image. Even worse is when that image is printed with those clipped highlights. I do realize we are talking about web posting, but most of the time that there is clipping going on (and I'm not talking about minor clipping here) it really jumps out right at me. My personal preference is to avoid them, whenever possible, save for a few exceptions (e.g. really deep shadows, eyes' pupils, silhouetted subjects, bright specular highlights, highlights reflecting off metal, etc...). I personally find clipped colours more "offending" than clipped whites or blacks.
Images are indeed judged by their own merit and not exclusively by the histogram, that is what art is all about! However, most of the images linked to could have been tweaked to avoid clipping and still would have won their respective accolades as it is the overall strenght of those images that captured the viewers in the first place. We are talking minute details perhaps, but that comes with the territory of always trying to improve personal results.
Sandra, in your case, it seems like you had very minor clipping going on and should be able to easily correct this during RAW conversion. I personal am happy with a very few "blinkies" on the histogram as I know I'll end up with a pretty darn good TIFF file once the RAW conversion has been done.
Fot those interested, there are a few fine threads dealing with this in the "Educational Resources" forum - some of which are linked here:
http://www.birdphotographers.net/for...Yellows-to-Hot
http://www.birdphotographers.net/for...9-REDucation-1
http://www.birdphotographers.net/for...1-REDucation-2
http://www.birdphotographers.net/for...y-and-HSL-INFO
John,
For the record, I never said ignore the histogram. What I said was in the process of producing an image to maker's expectation if histogram clips a bit ignore it.
I am simply sharing my opinion on the topic. As I repeatedly stated in pane #13. "We got our way of doing things and we should stick to it as long at it works".
Thanks
Sid
Last edited by Sid Garige; 10-18-2010 at 12:44 PM.
Hi Daniel, I linked those images because I am 100% certain they did not have clipping while in the larger color space and that the image reflect exactly the vision of the photographer and for me a perfect example why you do not need to adjust images that show clipping in the smaller color space of sRGB if the colors do not shift. Also we do not know if tweaking the images to fit the sRGB color space would have improved the image I have had images that loss their impact when tweaked to fit into the sRGB space and the reason I stopped worrying about it. Now when it comes to avoiding clipping of colors during capture or in the larger color spaces i agree that 99% of the time it has a negative effect.I guess I am on a campaign against the sRGB color space as a tool to view images and everyone should go out and buy a wide gamut monitor and use a color aware browser so it can fade into the dustbin of history:Dmost of the images linked to could have been tweaked to avoid clipping and still would have won their respective accolades as it is the overall strenght of those images that captured the viewers in the first place. We are talking minute details perhaps, but that comes with the territory of always trying to improve personal results.
Don Lacy
You don't take a photograph, you make it - Ansel Adams
There are no rules for good photographs, there are only good photographs - Ansel Adams
http://www.witnessnature.net/
https://500px.com/lacy
No problem Sid. I used the word to mean "earnestly urge" (OED). No offense was implied. Lots of good stuff here. I think we are getting to the bottom of this.
One issue about clipping not mentioned is that how offensive or offending it is depends partly on how the clipped pixels are distributed, and how many there are. If you had a largely white scene where 10% of the pixels were clipped at the high end, so long as they were distributed in amongst the detail, and contributed to it, you probably wouldn't notice them. If they were all clumped in one patch occupying 10% of the image you would.
I just find obvious patches of clipping in an image to be plain nasty in many cases, because ultimately it doesn't reflect what I saw with my eyes.
Last edited by John Chardine; 10-18-2010 at 01:16 PM. Reason: added qualifier
Don:
I think you are spot on here!
Norm
Very well summed it up Don.
Posting Permissions
Reply With Quote
