DenisV
newbie
Reged: 06/04/2008
Posts: 6
|
|
In the last issue of PM (12 April 2008) on page 29 Barney Britton replies to a question "Range roving" about why in some cases hystogram may be clipped at the high pont.
The answer Barney Britton provides is wrong or at least the answer is not there.
He writes about clipping darks/blacks or highlights. When this happens, the hystogram is skewed left or right repsectively and the clipped levels are represented by those vertical hystogram lines that do not fit within the range i.e. balcks do not fit within the left hystogram borger (brightness of below zero) and whites that do not fit withing the right hystogram border (brightness higher than 255).
On the horisontal axis the hystogram represents the level of brightness (from 0 on the left to 255 on the rigt). The vertical axis represents the quantity of pixels in an image that are of a particular brightness. The question of Leslie Kinder was about histogram "clippng" on the vertical axis. So I beleive the answer given is wrong or at best is not given or misleading.
If a hystogram is "clipping" at the top all this means is that there are a lot of pixels in teh image at this particular brightness level and the current scale of hystogram is not able to show this very tall line in full. There is nothing wrong with this as it does not really matter how many pixels are of a particular level in an image. in its most extreme case - a picture of an evenly lit surface without any brightness variation - the hystogram will be a single vertical line represnting all the pixels of hte image (nothing to the left or right of it). the picture itself will have no detail and will look the same as e.g. a computer screen showing nothng but e.g. white or grey.
A real world image with 'normal' exposuere on a hystogram will have no clipped blacks or whites (i.e. hystogram will fit well within the range horizontally) but the hystogram may or may not be "clipped" vertically. Likewise, for example an overexposed image will not fit within the right edge of the hystogram, but the fact if it does fit or not veritcally has nothing to do with overexposure.
Mr Britton in his answer implies not fitting within dynamic range (which is the horisontal scale of a hystogram) as the reason for clipping of vertical hystogram lines, without telling that by definiton of hystogram the vertical scale in fact represents the quantity of pixels of a particular brightness, not their brightness.
Denis
Edited by DenisV (06/04/2008 17:47)
|
barney britton
AP Technical Writer
Reged: 15/01/2007
Posts: 111
|
|
When Leslie wrote about the histogram clipping at the 'high points' he meant highlight areas of his photograph, and the highlight end of the histogram. My answer to his query was therefore correct. The histogram represents the tonal spread of an image, and when the majority of the data lies to one or other side of the median, this represents a prevalence of light or dark tones in the photograph.
Clipping is illustrated by data bunching up at the right or left hand sides of the histogram. A high line that 'runs off' either end represents data that is either pure black or pure white - either way, detail that is 'clipped' and cannot be recovered.
|
DenisV
newbie
Reged: 06/04/2008
Posts: 6
|
|
My understanding of Leslie's question was that it is about "clipping" of hystorgam at the top. Clipping of highlights, on the right, has been written about extensively everywhere and I do not think Leslie was on about it but rather wanted to see if "clipping" at the top meant anything and if it was important.
If only we could check with Leslie...
If he meant clipping highlights, I will have to apologise to Barney Britton.
Denis
|
Benchista
Wich Tyler
Reged: 11/08/2000
Posts: 37028
Loc: Everywhere and nowhere, baby
|
|
I would assume that Barney already has...
--------------------
Nick
www.nbrphoto.com
Light and Shade II - the new blog
|
TheFatControlleR
L'éminence Grise, Devil's Advocate & AP Fanboy!
Reged: 24/04/2001
Posts: 10841
Loc: In A Glass Darkly
|
|
I suspect there may have been more to the query than appeared in the magazine (edited for space/brevity) and the response was in keeping with the information given. IMO it's a pointless exercise trying to second guess when one isn't in possession of all the details.
--------------------
TheFatControlleR
And so, my fellow Benchists: ask not what your bench can do for you - ask what you can do for your bench... Ben Chism
Live fat, die young, and leave a self-basting corpse...
|
DenisV
newbie
Reged: 06/04/2008
Posts: 6
|
|
Gents,
No second guessing here. I read the question, I read the answer. The question is about why hystogram may be clipped at the top. The answer is about clipping darks/highlights (hystogram left and right). There is a gap.
Or am I wrong?
Denis
|
Benchista
Wich Tyler
Reged: 11/08/2000
Posts: 37028
Loc: Everywhere and nowhere, baby
|
|
You read the printed portion of the question - as Myk said, they're generally edited to fit the space. Given Barney is so insistant his interpretation is right, I would be surprised if there's not something in the rest of the original question that makes it clear that he is indeed right.
--------------------
Nick
www.nbrphoto.com
Light and Shade II - the new blog
|
Bettina
Kein Titel
Reged: 12/02/2004
Posts: 4060
Loc: London
|
|
... and it's "histogram" 
--------------------
Bettina
http://www.vibrantpictures.co.uk
Vibrant Pictures on Flickr
|
Angela Nicholson
AP Technical Editor
Reged: 07/02/2006
Posts: 142
|
|
The question is:
I have a Canon PowerShot G9 and shoot in raw, most often at low ISO settings. I always have the histogram turned on to assist my exposure settings. In some situations I find a histogram may be clipped at the high points. Why is this?
The term 'clipping' is usually used in association with the loss of highlights. The 'high points' could be taken either as the top of the peaks in the histogram, or the high (numerically)brightness values. As the terms 'clipping' and 'high points' were used in the same question the logical conclusion is that the question is about the likely reasons for overexposure.
Angela
|
huwevans
Old Hand
Reged: 05/08/2000
Posts: 15272
Loc: Dorset, UK
|
|
Indeed - I'd have to say (speaking as a mathematician) that that was to me by far the most rational interpretation of the question printed. So I think the poster here is just perhaps taking the reference to 'high points' in a very literal way, which isn't necessarily appropriate for the way that most people express themselves most of the time.
On top of that, I could be wrong but I thought most cameras when displaying the histograms autoscale the vertical axis. If that's the case then it should be impossible to have a situation when the data is clipped vertically. At any rate I can't recall ever seeing one apparently clipped in that way.
--------------------
Huw Evans.
  
|
DenisV
newbie
Reged: 06/04/2008
Posts: 6
|
|
Mr Britton,
Thank you, that is a sensible reply and it makes it a dialog.
Indeed, what I mean is that, if read directly, the reader asked about "high points" of histogram i.e. the top of the histogram, on vertical axis. Trying to be unbiased by how most other people would ask the question (and by how most experts would understand the question) being a reader of AP (and maybe of other magazines), same as me, he may have read about clipping highlights many times before and got that idea right by now. However, he (and me for that matter) may have never seen anything written about the vertical axis clipping on a photo histogram. I thought that was what he was on about.
Why not mention that aspect - what does the vertical axis mean and what if it is "clipped"? The number of pixels of a particular brightness (by definition), number too high to fit the line(s) in. Mention possibility of auto-scaling-to-fit. (BTW - I am not sure if Canon cameras auto-scale, at least I do not think 20D?). It might have been clear then and definitely make the answer more complete. There are two dimensions on a histogram - vertical and horisontal. Why write about horisontal here (and all the time) and not about vertical, especially when the question is about it?
So, was it about it? When I read it, I though it was. When I read it again – same.
If we look at the term "clipping", in its very basic form it is illustrated by a two dimensional line graph, where horisontal axis would be e.g. time and vertical will be e.g. signal level. Vertical scale limits can be dynamic range. Clipping, in this case, will be nothing other than the peaks of the graph line not fitting within the upper limit of the scale and therefore appear cut off at the top i.e. "clipped". Clipped at the top.
If we take a photo histogram, it is a different type of graph and it does not show clipping (of highlights) as such as in the case of line graph.
That means that the reader may well have meant "clipping" as in a line graph (cut offs at the top).
What do you think?
Can we ask the original poster?
Denis
p.s. Here are some examples of histograms clipped at the top.
Luminous Landscape - understanding histograms
Moderators Note: Link edited to fit forum
Edited by Fen (11/04/2008 12:44)
|
DenisV
newbie
Reged: 06/04/2008
Posts: 6
|
|
Dear Angela,
Your reply also makes sense. Indeed, there are two things here. "High points" (of histogram) and "highlights". The thing is, I thought that the conclusion that the reader meant that one is same as the other was a bit premature (there is nothing in the question to suggest that) and followed the easiest, well worn path - that of clipping highlights.
I understand the logic why the reply went as it did, I am not sure if the question was understood right. If the reader asked about the tops of histogram peaks, then the answer is whrong (it iwas about highlights) or at least incomplete (no mention of what vertical axis shows).
Or am I wrong?
Regards,
Denis
|
DenisV
newbie
Reged: 06/04/2008
Posts: 6
|
|
Here is another page I came accross when looking for examples of histograms clipped at the top. It turned out to be quite a comprehensive illustaration of photo histograms variations.
http://500th.net/histogram
There are even illustrations of the case when "a histogram is clipped at the top" - large areas of image without brightness variations, a lot of pixels of same (or almost same) brighntess level. Which is different from highlight cliping in more ways than one:
- different in how the histogram looks
- different in where they do not fit (vertical vs horisontal range)
- different in where clipping actually occurs (in highlights/darks case it is actual clipping of signal (not being able to record data representing actual level) due to limited dynamica rage while in case of histogram clipping at top it is simply that the scale chosen does not show the full lenght of the line counting pixels.
- different in what it means / should you worry about it (clipping highlights - lost unrecoverable data i.e. consider changing exposure; histogram clipping vertically - just illustrates in a certain way what kind of image has been recorded).
Denis
Edited by DenisV (11/04/2008 15:31)
|
MickLL
Two Grand Man
Reged: 01/08/2004
Posts: 2509
Loc: SE England
|
|
Come back RobertJ. 
MickLL
|
TheFatControlleR
L'éminence Grise, Devil's Advocate & AP Fanboy!
Reged: 24/04/2001
Posts: 10841
Loc: In A Glass Darkly
|
|
Innit.
----------------------------------------
I think we'll call it a day on this one.
Seems to me no amount of reasoning or explanation (even by the man who wrote the piece) is going to be accepted, and all that's happening is the re-stating of what has already been said.
|
Previous
Index
Next
Threaded
Edit
Reply
Quote
