FAQ's on HISTOGRAM [ Work in Progress ]

[mangaloreaviators]

Hello All,

I have been reading about Histogram in the last few days. Thought of collating all the Tutorials and links which explains about Histogram.

http://www.cambridgeincolour.com/tutorials/histograms1.htm (http://www.cambridgeincolour.com/tutorials/histograms1.htm)
http://www.luminous-landscape.com/tutorials/understanding-series/understanding-histograms.shtml (http://www.luminous-landscape.com/tutorials/understanding-series/understanding-histograms.shtml)

Please contribute

[PhotosGuy]

A PRACTICAL GUIDE TO INTERPRETING RGB HISTOGRAMS (http://www.sphoto.com/techinfo/histograms/histograms.htm)

Understanding Histograms (http://www.luminous-landscape.com/tutorials/understanding-series/understanding-histograms.shtml)

Towel method: Interpreting the Camera Histogram (http://super.nova.org/DPR/Histogram/)

Image Histogram Tutorial - Tones and Contrast (http://www.cambridgeincolour.com/tutorials/histograms1.htm)

Another simple explanation: Reading the Histogram (http://www.workshopsforphotographers.com/photo-imaging-tips-techniques/reading-histogram)

Advanced reading:

Curtis N found that a blown red channel is a problem:
How NOT to expose to the right (http://photography-on-the.net/forum/showthread.php?t=93712)
How to fix it:
Restore Those Clipped Channels (http://www.luminous-landscape.com/tutorials/restore-clipped.shtml)

[tzalman]

http://www.ronbigelow.com/articles/histograms-1/histograms-1.htm
http://www.ronbigelow.com/articles/histograms-2/histograms-2.htm
http://www.ronbigelow.com/articles/camera-histogram/camera-histogram.htm

[mangaloreaviators]

Hello PhotosGuy and Tzalman,

Thank you for the providing the links. The information in the link is very informative and I'm sure this will help the Newbies understand Histogram in detail.

@All other Visitors. Please submit all your questions related to Histogram here so that we can make this link really a FAQ's on HISTOGRAM.

Thank you for your time and keep posting :)

[mangaloreaviators]

Hello PhotosGuy and Tzalman,

Is there anything that we can do to make this thread more informative ?

[CAReefer]

Read through the links above as well as several pages gleaned from the search function and still haven't found an answer to my query. It may be a moot point, but I'm curious nonetheless.

Does the vertical height of the histogram come into play? More specifically, if the pixel count exceeds the limits of the visual graph? This seems to be where the majority of my pictures land. The first severe adverse affects I've noticed, were today after photographing some yellow roses in my DIY lightbox (thanks again pepperoni!). In checking the 'gram, I was within the horizontal limits and not clipping any of the shadows or highlights. Once in PP the yellow portion was significantly overexposed, well over 2 stops. Once dialed back in ACR 4.3, the rest of the image was pretty bad as you can well imagine. I did manage to do a reasonable job of saving it, only from a PS practice standpoint did I even waste the time rather than reshoot it since we won't always have that luxury.

Perhaps this is where one would be better served to use the RGB histogram if available? If so, what steps can be taken to get out of that range? I've tried EC, and going up a stop or two and all that does is shift it right or left, never dropping the pixel count.

So again, my question;
Is the vertical height (Pixel Count) critical once it exceeds the range of the histogram?

Thanks

[PhotosGuy]

Perhaps this is where one would be better served to use the RGB histogram if available? It can help to avoid clipped channel problems:
How NOT to expose to the right (http://photography-on-the.net/forum/showthread.php?t=93712)

More info on how correcting WB affects exposure levels in a bad way:
White Balance, the Secret Weapon (http://www.ppmag.com/web-exclusives/2007/11/white-balance-the-secret-weapo-1.html) I've tried EC, and going up a stop or two and all that does is shift it right or left, never dropping the pixel count. Generally, that's the way it's supposed to work. But I don't use EC: Need an exposure crutch? (http://photography-on-the.net/forum/showthread.php?t=89123)

BEGINNERS, STOP HERE! You've seen everything you need to know to start using WB correctly. Go through the links already posted again if something wasn't clear the first time. (Trust me on this). ;)

More on how the subject affects the exposure in Post # 47 (http://photography-on-the.net/forum/showpost.php?p=5191658&postcount=47)

[tdodd]

More info on how correcting WB affects exposure levels in a bad way:
White Balance, the Secret Weapon (http://www.ppmag.com/web-exclusives/2007/11/white-balance-the-secret-weapo-1.html)
If I understand what this article is saying correctly I do not agree with it at all.

Whether you shoot raw or jpeg, the sensor records RGB levels for each pixel and if there is too much of any one colour it blows that channel. The sensor doesn't care about colour temperature - if there's a lot of red/orange/yellow light then the red/green channels will blow first and the blue channel will be left floundering. That's the reality of the scene. Changing "white balance" is simply a piece of metadata that tells the software interpreting the raw data how to remap the values to generate a "balanced" output image.

If you want to accurately judge your RGB histogram, to make sure you haven't lost/clipped data, you don't want to be masking the true colour balance of the scene by trying to neutralise out a predominance of yellow, blue or anything else. The RGB histogram on the back of the camera is generated from the embedded jpeg thumbnail. If you shoot in tungsten light with a tungsten WB setting you are going to completely skew the impression of the actual balance of light in the scene.

When you shoot "raw" you are shooting to capture data, not to create an image. That is why you expose to the right - to maximise data capture. A lot of raw shooters use only the neutral picture style, because they do not want the camera fiddling around with saturation, contrast, colour tone etc. before generating the RGB histogram from which exposure judgements will be made. Well in my humble opinion, if you want to take that to its logical conclusion, you don't want the camera mucking about with the colour balance either, and telling bare faced lies about the actual RGB content as seen by the sensor.

For quite a while now I've been shooting in raw with daylight balance, regardless of the lighting conditions. I'll fix up my WB in post processing. Christ, it only takes a click or two to do, and certainly won't be worse than using some arbitrary preset in the camera. In fact, as of today, I've changed from a daylight setting to a fixed Kelvin value, because Lightroom doesn't seem to have the first clue what a daylight colour temperature should be from a Canon camera.

What I would quite like is some information from Canon on which Kelvin value is truly the most WB agnostic, so that I can refine my approach to generating RGB histograms to the most accurate possible. I'm guessing it might be somewhere between 5,200K (assumed as the temperature for the faithful picture style) and 5,500K which will do for "daylight".

The article, which talks about the exposure decreasing when you correct WB after the event, is completely missing the point. If your sensor was already maxed out on the red channel then you could not safely increase the exposure further. Fiddling with WB settings in the camera won't alter the fact that the sensor is capturing a predominance of red/yellow. You can lie to the camera if you want, and pretend there is less red/yellow, but that won't help you get a better raw capture. You'll either blow the red channnel or end up with a lower exposure in any case. The only way to compensate for a predominance of red/orange/yellow, at the time of capture, is the add more blue light to the scene in the first place.

Maybe I've got entirely the wrong end of the stick but I think not. I'm very happy to be corrected if I am wrong though.

If I'm right, this causes a bit of a quandry for the jpeg shooters. Do they want accurate RGB histograms or correctly white balanced images? A tough call to make. I'll just carry on shooting raw.

EDIT : By the way, I do understand that in ACR or any other raw software the RGB histogram is generated from the raw data rather than the embedded jpeg. But any adjustments are still just to tweak a piece of metadata - they do not affect the individual RGB values of the captured raw data. The only thing that will change the captured RGB raw data is changing the exposure, or changing the lighting on the scene.

[E-K]

Read through the links above as well as several pages gleaned from the search function and still haven't found an answer to my query. It may be a moot point, but I'm curious nonetheless.

Does the vertical height of the histogram come into play? More specifically, if the pixel count exceeds the limits of the visual graph? This seems to be where the majority of my pictures land. The first severe adverse affects I've noticed, were today after photographing some yellow roses in my DIY lightbox (thanks again pepperoni!). In checking the 'gram, I was within the horizontal limits and not clipping any of the shadows or highlights. Once in PP the yellow portion was significantly overexposed, well over 2 stops. Once dialed back in ACR 4.3, the rest of the image was pretty bad as you can well imagine. I did manage to do a reasonable job of saving it, only from a PS practice standpoint did I even waste the time rather than reshoot it since we won't always have that luxury.

Perhaps this is where one would be better served to use the RGB histogram if available? If so, what steps can be taken to get out of that range? I've tried EC, and going up a stop or two and all that does is shift it right or left, never dropping the pixel count.

So again, my question;
Is the vertical height (Pixel Count) critical once it exceeds the range of the histogram?

Thanks

A histogram is just a count (how high in the vertical) of the number of pixels at a given luminosity (the horizontal). The vertical is important with respect to allowing you to judge were things should fall. Going off the top will NOT result in and of itself in an overexpose image. Only going off the right end of the histogram will do that.

The issue with histograms is that if a relatively small number of pixels are overexposed than the actual count in the histogram is low and this may not be visible on the small histograms on the camera.

Maybe manufacturers will come up with a "spot" histogram where the histogram will only be for the pixels in a zoomed in area - or at least allow you to zoom in on the end of the histogram ;).

e-k

[E-K]

If I understand what this article is saying correctly I do not agree with it at all.

I would have to agree with your assessment. The author indicates correctly that adjusting the WB in RAW can in fact introduce more noise. However, there is an implication that setting the WB correctly will somehow affect your exposure. This would be patently untrue if basing your exposure off of a meter reading. It could impact you to some degree if chimping the histogram if your white balance is way off but the author made not mention that this is how she determined the original exposure.

If you want to accurately judge your RGB histogram, to make sure you haven't lost/clipped data, you don't want to be masking the true colour balance of the scene by trying to neutralise out a predominance of yellow, blue or anything else. The RGB histogram on the back of the camera is generated from the embedded jpeg thumbnail. If you shoot in tungsten light with a tungsten WB setting you are going to completely skew the impression of the actual balance of light in the scene.
...
What I would quite like is some information from Canon on which Kelvin value is truly the most WB agnostic, so that I can refine my approach to generating RGB histograms to the most accurate possible. I'm guessing it might be somewhere between 5,200K (assumed as the temperature for the faithful picture style) and 5,500K which will do for "daylight".


This article may help if you speak spanish...http://www.guillermoluijk.com/article/uniwb/index.htm

or if not do a google search on: raw histogram neutral channel multipliers, and you should come up with a few discussions.

e-k

[tdodd]

do a google search on: raw histogram neutral channel multipliers, and you should come up with a few discussions.

e-k
Thanks. I'm amazed I have not come across such a discussion in all the many months I have been on POTN. I suppose it was naive of me to think I was the first person ever to come up with this theory.

Anyway, after a quick Google, I've started reading this discussion....

http://luminous-landscape.com/forum/index.php?showtopic=22250&st=0

and hope to learn some useful things :)

Courtesy of Google, I've now come across such a discussion on POTN - http://photography-on-the.net/forum/showthread.php?t=442096

[René Damkot]

If I understand what this article is saying correctly I do not agree with it at all.

Not sure...

I *think* the article mentions "effective exposure" because she's looking at the histogram for judging the exposure of the image.
That's obviously influenced by the WB setting.

I agree, I think, that if no channels are clipped, WB setting shouldn't matter for exposure, unless you're using the histogram for judging exposure...

However, I've noticed that adjusting WB in LR one way or another does influence the clipping of channels...

So; what happens when WB is set in the raw converter? Is one channel (the highest?) pushed down, or another pulled up? Or both?
If a channel is pulled up, I can imagine clipping and additional noise can occur.


Oh boy, more articles to read ;)

[JeffreyG]

I think I get it.

WB setting has no impact on the acrual RAW capture...and so the WB you select will not affect the outcome of a blown color channel in any meaningful way.

BUT.....

The WB will affect how the RGB histogram displays. So if you are shooting in a feedback mode from the histogram (and if you are exposing to the right then you probably are) using the wrong WB can lead you to make bad decisions about the color exposure.

Is this now correct?

[tdodd]

JeffreyG - You've got it :) The problem is, we don't know which WB setting we should be using in order to obtain an accurate and unbiased RGB histogram from the embedded jpeg. If you Google for uni-wb you will see that in the Nikon camp people have created a special custom white balance setting to achieve this "neutral" WB setting. A different file is required for each camera model.

So far, on the Canon side of things, I am not convinced by what I have seen - so I'm sticking with a middling WB of 5,500K for now, regardless of what I shoot. I have seen sample CR2 files for the 30D and 40D, for example, that are supposed to be used to set a white balance on the camera, but the approach to their creation seems highly suspect to me and their appearance looks decidedly odd and unconvincing. You can find the sample Canon uni-WB files on this page - http://www.guillermoluijk.com/article/uniwb/index.htm - near the bottom, in this section "DESCARGA DE ARCHIVOS RAW UNIWB".

I've attached a screen print of the thumbnails of the "calibration" target images for several cameras. The variance in them seems very extreme and I fail to appreciate the reason for some of the files to show graduated shades - surely they should be one solid colour. I think they are actually all photographs of a PC monitor, and as such, we are seeing the problem of angular falloff from the LCD displays. This makes them useless, as far as I am concerned, for any sort of calibration. To make matters worse, it has been said in the discussions supporting these images that screen calibration is not important when producing the images on screen to be photographed. Well, unless I've missed something in the theory behind the creation of these images, that just seems like nonsense to me.

I totally agree with the principle of establishing a uni-WB calibration target for our cameras but I just don't think the execution is right in these examples. I also think some of the theory that has been used may be flawed with respect to the green channel and baselining it differently from red and blue. There may be twice as many green receptors on a Bayer chip than red or blue, but that does not mean the green channel will blow any quicker than the others, so I am not sure why any special consideration has been given to green.

It would be so simple if Canon would make available uni-WB CR2 files for each body.

[E-K]

So far, on the Canon side of things, I am not convinced by what I have seen - so I'm sticking with a middling WB of 5,500K for now, regardless of what I shoot.

Using a colour temperature of 5500K on my Rebel XT results in a WB Adj RGB levels of (1024, 1024, 1443).


I have seen sample CR2 files for the 30D and 40D, for example, that are supposed to be used to set a white balance on the camera, but the approach to their creation seems highly suspect to me and their appearance looks decidedly odd and unconvincing. You can find the sample Canon uni-WB files on this page - http://www.guillermoluijk.com/article/uniwb/index.htm - near the bottom, in this section "DESCARGA DE ARCHIVOS RAW UNIWB".

Have you tried them? I tried the 350D one using DPP. It resulted in WB Adj RGB levels of (1024, 1024, 1035). Not perfect but better.

I've attached a screen print of the thumbnails of the "calibration" target images for several cameras. The variance in them seems very extreme and I fail to appreciate the reason for some of the files to show graduated shades - surely they should be one solid colour.

I think they are actually all photographs of a PC monitor, and as such, we are seeing the problem of angular falloff from the LCD displays. This makes them useless, as far as I am concerned, for any sort of calibration.

No, as long as the ratio between RGGB values is the same then it shouldn't matter. You are using it for WB not exposure.

To make matters worse, it has been said in the discussions supporting these images that screen calibration is not important when producing the images on screen to be photographed. Well, unless I've missed something in the theory behind the creation of these images, that just seems like nonsense to me.

In the discussion the bolded part is important. The screen doesn't need to be calibrated, but your shot of the gradiant image and your shot of the resultant solid color have to be on the same monitor and in the same lighting conditions. It's essentially the same principle that allows custom WB to work in the first place.

e-k

[tzalman]

There is a very interesting free application called Rawnalyze (http://www.cryptobola.com/PhotoBola/Rawnalyze.htm) that will show you, among other things, a histogram of a RAW file and also a histogram of the image with WB applied and mapped to sRGB space. The WB is determined by selecting an area of neutral grey and hitting control+W. Here are some captures from its UI. They are histograms of shots of this grey card.

http://photos.imageevent.com/elied/linkto/_MG_2226.JPG

The first one is exposed at +1/3 EV from the meter reading. The light was daylight from an overcast but bright sky. Note that the green channel is more exposed than the other two. This is because the CFA sensor is more sensitive to green than to red or blue so in most lighting which contains the entire visible spectrum the relationship between the channels will be similar although differing somewhat according to the color of the light. (Exceptions are where the light is nearly monochromatic, for instance, light reflected from an object whose color is very saturated, like a rose, will have almost no green or blue)

http://photos.imageevent.com/elied/linkto/Raw%20histo%201.jpg

The second image is after WB is applied in sRGB space. The WB is done by holding the channel with the highest value at 1.0 (usually the green) and multiplying the other two so that in the selected grey area they will be equal. The coefficients are listed above the histogram.

http://photos.imageevent.com/elied/linkto/mapped%20histo.jpg

The third shot is exposed to the right.

http://photos.imageevent.com/elied/linkto/Raw%20hist%202.jpg

And this is with WB. Note that the green is unchanged but the red and blue have clipped.

http://photos.imageevent.com/elied/linkto/mapped%20histo%202.jpg

Now, what does the camera LCD show us for this ETTR image? First with WB set to Auto. All the channels appear clipped althousgh we have seen that they are not.

http://photos.imageevent.com/elied/linkto/AWB.jpg

And now with the UniWB for the 40D as downloaded from Guillermo's site (and made, BTW, by the author of Rawnalyze, Gabor Schorr, aka panopeeper) The histogram resembles the RAW histogram because it is really a non-WB. Since the coefficients are all 1.0 the image remains unbalanced, just like the RAW is.

http://photos.imageevent.com/elied/linkto/UniWB.jpg

[tdodd]

E-K, obviously you know a lot more about the topic than I do. But how do you explain that the target images for the 30D shows a gradation from pinkish at the top to green/blue - maybe cyan - at the bottom? Why are some of the sample images a solid tone, which I would expect, while others seem to have a relatively broad spread through the colour spectrum? The "problem" that I see is not simply about the variation in brightness of the sample image. Quite clearly (to my eyes) the colour is inconsistent from top to bottom. This does not make sense to me.

I don't understand your measurements of relative RGGB values. Are you using DCRaw? It won't even install on my laptop - I'm using Vista X64.

Clearly my lack of Spanish is letting me down and the discussion thread I found has not explained the procedure in terms which I understand. Oh well :(

Why are the WB calibration targets virtually plain grey for the 40D and yet varying shades of pink/magenta for the other cameras? Something here does not add up in my tiny brain. I will try again to get my head round this but at the moment I am struggling. If I can't measure my raw RGGB values then it seems I'm up a gum tree.

[tdodd]

tzalman, thanks for your help and guidance. I applied the uni-WB file for my custom white balance and tried to follow your process pattern. I can see the results are heading in the right direction but something is not quite right. Maybe it is my mistake or maybe I need a different WB uni-WB profile for my 40D. Here are my results....

Firstly I took a shot of a "white" ceiling, in daylight conditions, but obviously in its own shade. I shot this simply to establish a baseline for judging how the RGB histogram looks under "normal" conditions of white balance. The camera was on a tripod for exact repeatability of the second shot, with the custom white balance.

Here is the first shot, using AWB....

http://farm3.static.flickr.com/2115/2401215660_1d4d282ce4_m.jpg

Here is the histogram on the back of the camera, showing a very even distribution of RGB, as one would expect....

http://farm4.static.flickr.com/3184/2400383979_d5d31c61c9_m.jpg

Here is the same shot, this time with the custom uni-WB....

http://farm4.static.flickr.com/3007/2401215624_3ba2c41854_m.jpg

and the histogram on the back of the camera for that shot. As one would expect, the RGB histogram now shows a significant emphasis on green and red has retreated a long way back. Who would think that an ostensibly "grey" custom WB image would have such a marked effect on the WB within the camera?!...

http://farm4.static.flickr.com/3149/2400384095_56f124e91d_m.jpg

Now for the analysis in the raw image analysis software. Here is the original raw data, unmapped to any particular colour space. As you can see, the histogram bears some resemblance to that from the back of the camera. Green is clearly in the lead and red is trailing at the back.

http://farm4.static.flickr.com/3255/2400384273_97c4fa6a3f_m.jpg

Here is the raw data mapped to the sRGB colour space. Once again, we see green out in front and red trailing at the rear....

http://farm4.static.flickr.com/3210/2401215768_4953dfcd48_m.jpg

And finally, here is the raw image data when mapped to sRGB and white balanced within the raw analysis software....

http://farm3.static.flickr.com/2041/2401215786_df7fac4439_m.jpg

I don't know which of the first two histograms from the software should be matched to the RGB histogram from the camera, but either way, while they share some traits in common, there is still quite a discrepancy. For example, on the back of the camera there is a clear gap between the leading edge of red and the trailing edge of green. But in the histograms from the raw analysis software there is a distinct overlap between the bright reds and the dim greens. So it seems to me that this WB calibration is still not quite correct. Maybe the only really important thing is that we identify highlight clipping points accurately, but it strikes me that with this calibration the camera could indicate little or no clipping in the red channel, when in fact there was a lot of clipping going on.

I'd be grateful for any further input from the experts on this :)

[tdodd]

In addition to my confusion above, I don't understand this suggestion that the sensor is more sensitive to green than red or blue. Certainly there are twice as many receptors for green, but that just means twice as much light will be captured in the green channel. Surely each green receptor is no more sensitive to light than any of the red or blue ones. That being the case, I still don't see why the theory of uni-WB has to emphasise greens in the histogram - there may be more of them but none of them will blow out any quicker than any other channel. I think this whole green thing is a complete red herring.

If the histogram truly represents the raw data then there should be twice as many green pixels than red or blue in the histogram. The green measurements might go higher in the histogram to reflect their greater quantity, but why should that cause the green pixels to also be pushed to the right in the histogram. Just because there are more of them does not mean they are any brighter or their wells more full.

Can someone please explain why a special case is being made for green in the histogram? Maybe it's all there in the Spanish article, but as I don't understand Spanish...... :(

[E-K]

E-K, obviously you know a lot more about the topic than I do. But how do you explain that the target images for the 30D shows a gradation from pinkish at the top to green/blue - maybe cyan - at the bottom? Why are some of the sample images a solid tone, which I would expect, while others seem to have a relatively broad spread through the colour spectrum? The "problem" that I see is not simply about the variation in brightness of the sample image. Quite clearly (to my eyes) the colour is inconsistent from top to bottom. This does not make sense to me.

Some of the provided samples are not the best (they tend to have a larger error as well). The 30D one seems to be overexposed on the bottom (LCD with the backlight at the bottom?). None the less, the Custom WB uses the central portion of the image and in that area, the ratio is pretty consistent.

I don't understand your measurements of relative RGGB values. Are you using DCRaw? It won't even install on my laptop - I'm using Vista X64.

I didn't measure the relative RGGB values, that was just a statement ;). To guesstimate for the analysis of the 30D I just used UFRaw and sample a few areas. You could use the RAWAnalyze tool that tzalman mentioned as well.

The WB Adj RGB levels I pulled out of the RAW file using EXIFTOOL.

What I would suggest is using the appropriate file as a source for your CWB, take a picture and then look at the WB Adj RGB or WB RGGB Levels As Shot setting from ExifTool. It should be very similar values for the three numbers.

Why are the WB calibration targets virtually plain grey for the 40D and yet varying shades of pink/magenta for the other cameras?

To some extent the same reason that a shot in daylight with a tungsten WB looks blue (i.e. your looking at the image after WB has been applied). The other reason would be variations in the camera itself.

e-k

[tdodd]

Thanks, E-K. Before I continue with this I'm going to have to find answers to my concerns about this green channel thing, as I described in post #19 above.

[tzalman]

tdodd - The difference between the first two Rawnalyze histograms is because the mapping to sRGB reqires a gamma 2.2correction. Since the camera's jpg is also gamma corrected, I suppose it should most resemble the second histogram. However, I have no idea how accurate this DIY UniWB target is, i.e., if it really triggers absolutely no white balancing. I think we shouldn't expect it to be more than an aid. Moreover, since the circumstances when we would be looking for guidence from it are are those in which we are on the edge of clipping, surely that is the sort of exposure we should be testing. I do take your point about the possibility of its failing to show red clipping. I think tomorrow I will look for a highly saturated red subject, shoot a series of exposures until it is clearly clipped and do some comparisons.

[tdodd]

I'm going to re-read whatever I can find in English on this topic and try and work through it to create my own uni-WB. If I'm successful I'll happily share it with people. I'm not convinced from what I've seen that we're at the end of the journey on this one yet.

I ran my uni-WB balanced file through Exiftool and have captured a small screen print of the RGGB values hard coded into the raw file, and those variable RGGB values created from the uni-WB balance. Certainly the RGGB values for the custom WB are the closest to each other, but is that the goal? Are they supposed to be identical in order to get the best uni-WB calibration we can get? I'm struggling to get to grips with that one. Do we know what those values actually mean in Canon's world? How do we know, or why do we think, that the RGGB values should match. Who says they should? Do they directly represent the multipliers established in the raw image analysis software?

Too many questions at the moment and too few answers - for me at least.

EDIT : I've been pondering the values in the listing attached below and I'm thinking that the baseline RGGB value we should be aiming for is more like 2048, 1024, 1024, 2048. If you think about it, there are twice as many green doobries and most of the GG values are 1024, 1024 or very close, totalling 2048 thingies for the GG channel. If you look at the R and B values (ignoring the custom uni-WB stuff for now) R values range from 1653 to 2564 and B values range from 1245 to 2503. If these are some sort of multipliers then it is my guess that the "zero" point is around a level of 2048, not 1024.

Take the tungsten WB (correction) values for example. Now we know that to balance a tungsten environment we need to subtract reddish/yellowish tones and increase blueish tones. The RGGB values for tungsten are 1653, 1103, 1103, 2503. You see, the red channel has been reduced, not by increasing its multiplier from 1024 up to 1653 but rather by reducing it from 2048 to 1653. Similarly, the blue channel is at 2503, which is higher than 2048 and much much much higher than 1024.

This interpretation seems to back up my contention that the green channel is not more sensitive as such, it's simply that there are twice as many green sensors. In order for the red and blue channels to catch up, and give a picture with an evenly balanced appearance they each need their "strength" to be doubled. Hence my logic of 2048 for each of R and B and 2*1024=2048 for green.

Maybe I'm talking complete nonsense. Someone put me out of my misery.

[E-K]

I ran my uni-WB balanced file through Exiftool and have captured a small screen print of the RGGB values hard coded into the raw file, and those variable RGGB values created from the uni-WB balance. Certainly the RGGB values for the custom WB are the closest to each other, but is that the goal? Are they supposed to be identical in order to get the best uni-WB calibration we can get?

Yes that is the goal.

I'm struggling to get to grips with that one. Do we know what those values actually mean in Canon's world? How do we know, or why do we think, that the RGGB values should match. Who says they should? Do they directly represent the multipliers established in the raw image analysis software?

Yes, they correspond to the multipliers that would be used by a RAW converter which new how to get them out of the file.

In UFRaw for example (gives direct access to the multipliers), I can calculate the multiplier values manually by using the ExifTool info and dividing all of the values by the value for the green channel.

As to why you want the values equal: The idea is that, at least with an RGB histogram, by using a WB of (1, 1, 1), you get an accurate view of the information in each of the channels. Otherwise you may see soft clipping when there really isn't any hard clipping occurring.

This is primarily a benefit when trying to expose to the right.

e-k

[partsman]

Clearly my lack of Spanish is letting me down and the discussion thread I found has not explained the procedure in terms which I understand. Oh well

heres a link (http://translate.google.com/translate?u=http%3A%2F%2Fwww.guillermoluijk.com%2F article%2Funiwb%2Findex.htm&langpair=es%7Cen&hl=en&ie=UTF-8) to the artical translated into english.

[tdodd]

Thanks for the link. It's quite slow progress reading the translated text but hopefully I'll get there. Anyway, I got a bit distracted and went back to some practical work instead of research and had a go at generating my own uniWB for my 30D. Using ExifTool to monitor my measured RGB values I created a target colour using Windows Paint. I now understand why monitor calibration is quite irrelevant.

I've attached my Paint custom color settings for my final uniWB and the measured values from ExifTool. The figures in Paint seem to be very skewed. I had to dial in a massive amount of red in order to bring the measured RGB red values down to the level of the green channel, and I even had to lower the green channel quite a way. My monitor is calibrated to D65, quite recently.

Other than irregularities from the monitor being off colour, I would think that R and B values in Paint should be quite close, as their sensitivity is, I assume, quite close to each other. My uniWB colour does look somewhat like the original 30D calibration target in hue, so maybe I am on the right track, but can the 30D really be so insensitive to red light? The multipliers for red seem very extreme.

If this is the amount of correction required to generate an accurate RGB raw histogram from the embedded jpeg it's a wonder that with normal WB settings the RGB histogram is any use for anything. I can't help feeling there is some flaw somewhere. I guess I'd better go back to that translated article.

[tdodd]

I've now been exploring this subject at length and tried out ExifTool, RawAnalyze and UFRaw to assess my true raw histogram (not the interpretations from jpeg files) and have decided to go with the flow and completed calibrations pretty accurately for my 30D and 40D. Whether the effort proves worthwhile in practice remains to be seen, but at least I've given it everything I've got. By the way, I dumped UFRaw - when it came to mapping (or whatever the term is that UFraw uses) from the raw data to an sRGB target it indicated clipping where the raw histogram indicated otherwise. I thus limited my interest to ExifTool and RawAnalyze.

I've created a little online album to show my overall setup for the calibration task, my desktop environment for fine tuning the target and to upload a large/fine jpeg straight from camera for the final 30D and 40D calibration images. I performed both calibrations with the camera set to daylight white balance, so both targets look pink, so as not to cause confusion by having them captured under different WB conditions.

Here's the overall setup, with my LCD TV acting as the calibration target and my camera tripod mounted with my 100-400 lens and hood to minimise stray light contamination. Although it was not pitch black by any means, the room was darkened, as you can see from the drawn curtains.

http://lh6.ggpht.com/EezyTiger/R_9V_-Z_93I/AAAAAAAAlL8/DE4GduD1tCo/s800/20080411_115757_02866.JPG

Here's my working desktop. I used EOS Utility remote capture to fire the camera and analyse the resulting raw file with ExifTool to get the measured RGGB values as close to each other as I could. It took many many tweaks to my target colour on the TV until everything was lined up as well as I could manage. Once the results were good in ExifToool I used RawAnalyze to verify the raw histogram and RGB coefficients afer running the WB calibration in RawAnalyze. The results appear to be excellent. The four pink images you see in DPP are the final raw and jpeg files for the 30D and 40D.

http://lh6.ggpht.com/EezyTiger/R_9V1-Z_92I/AAAAAAAAlL0/vuXGOhlnJbE/s800/20080411_130525_002.jpg

The album with these two images and the target calibration jpeg files can be found here - http://picasaweb.google.co.uk/EezyTiger/CanonUniWB?authkey=AagygA3j5nM. The two calibration files are the full jpegs, straight from the camera, so unmolested in any way (except renaming them), and should be fine for calibration in another camera. You can download them from the website. The files have been named so that they can be copied straight to a CF card and used as the custom WB calibration images for each camera. File names are IMG_0030.jpg for the 30D and IMG_0040.jpg for the 40D.

[tdodd]

Maybe I'm expecting too much from this uniWB calibration but I've started trying to work through what advantage it will give me in the real world. I'm afraid I have already hit upon a stumbling block very early on in putting it to the test.

Firstly I wanted to test the accuracy of the RGB histogram reporting with some simple test colours, produced on my LCD TV. I know these colours may not be colour accurate, but if I want a predominance of red, or green, or blue it is easy to achieve. So I set the TV up to display an RGB = 0, 255, 0 image and took some test photos using manual exposure to set exposure beyond +2 stops if necessary. I am confused by the results.

Here is a screen print, showing data presented by different software from one image file.


http://lh3.ggpht.com/EezyTiger/SABy3-Z_-BI/AAAAAAAAlQY/A4hQ9kYMj4o/s800/MWSnap011.jpg

Link to larger screen print image here - http://picasaweb.google.co.uk/EezyTiger/CanonUniWB/photo?authkey=AagygA3j5nM#5188273076701231122

Remembering that the image was of a "pure" green display on my TV, and that the uniWB should make things look even greener than they should be, let's take each software display in turn....

Top left: Lightroom - The image does look a fairly rich green, which is what we should expect. The Develop view indicates severe highlight clipping in the green channel across a large portion of the image.

Top right and bottom right : DPP - The image looks washed out and very nondescript. Sampling various points within the image the highest value I can find for green is 244, which is well away from being clipped. In the bottom right we have the RGB histogram for the image. This clearly shows some space to the right of the green spike. There is no clipping identified anywhere, but R and B values seem high, and the image is looking a long way away from "pure" green and headind towards nothing more than green tinted grey/white. This suggests to me severe overexposure, with red and blue catching up to the green channel, which is already maxed out and can go no higher.

Bottom right, one window in : Camera histogram - We can see the predominance of green in the RGB histogram, but red and blue are quite close in luminance. The image on the camera looks extremely washed out, a lot like the display in DPP. In fact, the camera and DPP seem to have very much the same interpretation of the image.

Bottom centre : RawAnalyze - This is the raw histogram, unmapped to any colour space. Supposedly it is a true representation of the stimulus received by each RGB pixel. In this example the green channel is clipped - you should be able to see a spike on the extreme right that is only a single pixel wide on your screen - but the spike is there. Red and Blue are well over to the left, as I would expect for an image that was "pure" green.

Bottom left : This is the output from ExifTool. It confirms that the "as shot" RGGB multipliers are 1030 1024 1024 1034, which is very close to the ideal for uniWB.

So what has gone wrong? Why do Lightroom and RawAnalyze think the green channel is blown when the camera and DPP think otherwise? Why does Lightroom present a rich green image, as one would expect when the camera and DPP render a washed out image? It certainly looks like everything has gone pear shaped in the Canon camp and I cannot trust the uniWB calibration to yield meaningful information on the back of the camera.

Any suggestions? Where did I go wrong?

[tdodd]

Here's a similar set of data for a "pure" blue image. I left the exposure settings unaltered....

http://lh3.ggpht.com/EezyTiger/SAB-1-Z_-CI/AAAAAAAAlQ4/ZyCr0h4Emuo/s800/MWSnap012.jpg

Fulll screen print available here - http://picasaweb.google.co.uk/EezyTiger/CanonUniWB/photo?authkey=AagygA3j5nM#5188286236481026082

This time RawAnalyze does not consider any channels to be blown and, in terms of relative strength, blue is by far the brightest, while green leads red. This analysis is echoed in DPP with it's histogram and spot sampling the image. However, Lightrrom seems to think there is something blown in the blue channel - maybe just one pixel - but not enough to actually highlight which pixel(s) is/are blown. Although the image in Lightroom looks rich blue, remember that the perceived colour should be skewed towards green by the custom WB calibration, so to look "pure" blue does not seem right to me. Finally there is the issue that Lightrrom has reversed the dominance of red vs green. Why?

So once again things don't seem to add up. How is one to know which software to trust? It seems to me there are lies, damned lies, and RGB histograms.

[tdodd]

Finally, the same thing again for a "pure" red image. Once again, exposure settings were unaltered.

http://lh3.ggpht.com/EezyTiger/SAB-2-Z_-DI/AAAAAAAAlRA/EYVRGEHBn0Y/s800/MWSnap013.jpg

Full screen print image available here -
http://picasaweb.google.co.uk/EezyTiger/CanonUniWB/photo?authkey=AagygA3j5nM#5188286253660895282

This time, all the ducks seem to be lined up correctly, with red in the lead, trailed by green and then blue. The RawAnalyze histogram shows that red was a long way from being saturated, and Lightroom seems to agree. But now look at DPP. It thinks the reds are completely blown and it is easy to sample pixels with R=255 over the majority of centre of the image.

Across the three tests of red, green, blue images, if we assume that RawAnalyze is the oracle of truth, it seems that both Lightroom and DPP are completely clueless when it comes to converting that data to an image and generating a reliable histogram. As the camera display seems to closely resemble the displays from DPP I'm still as much on the dark as I ever was, in terms of maximising dynamic range captured in my raw files.

[E-K]

Maybe I'm expecting too much from this uniWB calibration but I've started trying to work through what advantage it will give me in the real world. I'm afraid I have already hit upon a stumbling block very early on in putting it to the test.

As an aside, I've looked into an easier way to create a UniWB image (see http://photography-on-the.net/forum/showthread.php?t=485349).

Firstly I wanted to test the accuracy of the RGB histogram reporting with some simple test colours, produced on my LCD TV. I know these colours may not be colour accurate, but if I want a predominance of red, or green, or blue it is easy to achieve. So I set the TV up to display an RGB = 0, 255, 0 image and took some test photos using manual exposure to set exposure beyond +2 stops if necessary. I am confused by the results.
...
So what has gone wrong? Why do Lightroom and RawAnalyze think the green channel is blown when the camera and DPP think otherwise? Why does Lightroom present a rich green image, as one would expect when the camera and DPP render a washed out image? It certainly looks like everything has gone pear shaped in the Canon camp and I cannot trust the uniWB calibration to yield meaningful information on the back of the camera.

Any suggestions? Where did I go wrong?

What are you using for your picture styles, contrast settings, etc... These all have an impact and are really only taken into consideration in DPP.

e-k

[tdodd]

Finally, finally, perhaps this changes everything and "proves" that for real world photography, with a full gamut of colour, uniWB does pretty much the best job possible.

This time I shot a "pure" white image on the TV screen. I metered in manual to obtain a centred reading from the meter in centre weighted average mode and fired off a shot. Here are the results....

http://lh3.ggpht.com/EezyTiger/SACM3-Z_-EI/AAAAAAAAlRg/_0Brs0BsTBM/s800/MWSnap014.jpg http://lh3.ggpht.com/EezyTiger/SACQ3-Z_-GI/AAAAAAAAlSI/wJD1PiquCe4/s800/DSC02873.JPG

Full sized screen print - http://picasaweb.google.co.uk/EezyTiger/CanonUniWB/photo?authkey=AagygA3j5nM#5188301664003553346

RawAnalyze shows I have almost three stops of headroom before I blow the green channel. All histograms match quite closely, with green leading blue leading red and a fair bit of overlap between them. The expected green tint is clearly visible in the "pure" white image, before correcting WB to a more normal value. It all seems to be working rather well :)

I bumped up the exposure by 3 stops and fired off another test image....

http://lh3.ggpht.com/EezyTiger/SACM4-Z_-FI/AAAAAAAAlRo/YZb36vEbCsk/s800/MWSnap015.jpg http://lh3.ggpht.com/EezyTiger/SACQ6-Z_-HI/AAAAAAAAlSQ/p1UDApw-2Ws/s800/DSC02872.JPG

Full screen print - http://picasaweb.google.co.uk/EezyTiger/CanonUniWB/photo?authkey=AagygA3j5nM#5188301681183422546

What can I say? This actually looks very promising indeed. Once again there is close alignment between all the histograms.

I will leave my WB set to uniWB and see how things go.

[tdodd]

As an aside, I've looked into an easier way to create a UniWB image (see http://photography-on-the.net/forum/showthread.php?t=485349).

Interesting. I'll give it a go.

What are you using for your picture styles, contrast settings, etc... These all have an impact and are really only taken into consideration in DPP.

e-k
Good point! I shoot with neutral picture style and all adjustments for sharpness, contrast, saturation and tone at defaults within the camera. That means no sharpening at all but with the other settings there is room to reduce them to negative values.

[mangaloreaviators]

Hello All,

i was reading this article from DP Review http://www.dpreview.com/learn/?/key=histogram

The article says all dark things will be there on the Left Side of Histogram and all light things will be there on the Right side of Histogram.

Correct me here.

If I take a picture of a Black Board the Histogram will be pushed towards the Left and if I take picture of a White Board the Histogram will be pushed toward the Right.

If the above is correct.
Then if I take a picture of a RED Board--It will be there somewhere between the middle and left of the Histogram Correct ?

[E-K]

If I take a picture of a Black Board the Histogram will be pushed towards the Left and if I take picture of a White Board the Histogram will be pushed toward the Right.

Assuming you have exposed to keep the black board black and the white board white then yes this is correct.

If the above is correct.
Then if I take a picture of a RED Board--It will be there somewhere between the middle and left of the Histogram Correct ?

Well it could be anywhere depending on how bright the red board was and what your exposure was. For example, if it is a really dark red it could be left of centre.

e-k

[tdodd]

To add a little more meat to E-K's reply....

With the camera in autoexposure mode it will try to set an exposure to average out the scene to a middle grey tone. Thus, if you photograph a solid black object the camera will calculate a long exposure so that it collects enough light to render the scene as grey. That will give you a histogram spike in the middle. If you shoot a white object the camera will calculate a very short exposure so that it only captures a little of the reflected light, again turning the scene into grey and giving you a spike roughly in the middle. The same thing will happen with any single solid tone/colour - the camera will try to make it an average intensity of brightness.

Now, if you were to shoot a scene that was 50% black and 50% white the camera would again try to create an average grey from the scene. The histogram would show two spikes - one on the left for the black subject amd one on the right for the white subject. There would be no spike in the middle. If you shot a scene of 33% black, 33% red and 33% white you would see three spikes across the histogram, representing the different respective brightnesses of each element of the scene. If you viewed an RGB histogram, rather than a plain brightness histogram, you would clearly see where the red peak sits.

Cameras are quite dumb when it comes to autoexposure - they just want to calculate an average grey across the scene. They don't know or care whether you are shooting a snow scene or a coal cellar. The skill of the photographer is to understand the scene and instruct the camera on how the scene should be represented, by using exposure compensation or simply shooting in manual exposure and setting the exposure to fit the scene.

[mangaloreaviators]

@EK - Will try some steps to check if that works.

Todd I tried the test to see if the Histogram goes to extreme Left if I shoot a Blackbody and towards the right if I shoot a White board. The Camera did the same thing that you explained.

" With the camera in autoexposure mode it will try to set an exposure to average out the scene to a middle grey tone. Thus, if you photograph a solid black object the camera will calculate a long exposure so that it collects enough light to render the scene as grey. That will give you a histogram spike in the middle. If you shoot a white object the camera will calculate a very short exposure so that it only captures a little of the reflected light, again turning the scene into grey and giving you a spike roughly in the middle. The same thing will happen with any single solid tone/colour - the camera will try to make it an average intensity of brightness."

So I switched the Camera to Manual Mode and placed the Camera vertical on a Black Board and clicked by looking through View Finder. The Histogram was pushed completely towards the Extreame Left side of Histogram.

When I did a similar procedure for a white board the image came grey. So I used the Exposure compensation to the Max and the image was towards the Right of Histogram

[Canon Soldier]

sticky?

[_GUI_]

you are having a lot of fun here with the UniWB, right? :)

I keep my camera continuously configured in UniWB. In fact in the little 350D the trick to shoot a totally blown RAW works perfect to set UniWB. 5D and several Nikon users have reported that this method does not work on their cameras. In fact if it works on the 350D is thanks to the bad implementation of its Custom WB routine, which considers as valid for the calculation saturated pixels.

BR

[E-K]

In fact in the little 350D the trick to shoot a totally blown RAW works perfect to set UniWB. 5D and several Nikon users have reported that this method does not work on their cameras. In fact if it works on the 350D is thanks to the bad implementation of its Custom WB routine, which considers as valid for the calculation saturated pixels.

Seems a little odd. I wouldn't have thought the camera WB logic would try to distinguish saturation from pure white. Did the 5D and Nikon users confirm that the raw data was in fact saturated? Did the dark frame method work for those cameras?

e-k

[_GUI_]

Seems a little odd. I wouldn't have thought the camera WB logic would try to distinguish saturation from pure white. Did the 5D and Nikon users confirm that the raw data was in fact saturated? Did the dark frame method work for those cameras?

e-k

I consider it very logical, think that any partially saturated pixel has incorrect R/G/B ratios, so using them to calculate white balance will surely lead to a wrong white balance. Any colour when saturated becomes white; this does not mean that any colour is neutral.

I didn't understand well how the dark frame method could work (there is nothing to balance on dark shots but noise, and there is no reason to think noise should be neutral) so I didn't propose it.

[E-K]

I consider it very logical, think that any partially saturated pixel has incorrect R/G/B ratios, so using them to calculate white balance will surely lead to a wrong white balance. Any colour when saturated becomes white; this does not mean that any colour is neutral.

What I find odd, is that if this was the case I would expect the camera to reject the image selected for WB outright as unsuitable.

So the question is was the WB affected at all when these saturated images were selected on the 5D. If so, then the camera accepted the image as a WB image. If the camera accepted the image, then I would hypothesize that the issue could be that all channels weren't fully saturated.

I didn't understand well how the dark frame method could work (there is nothing to balance on dark shots but noise, and there is no reason to think noise should be neutral) so I didn't propose it.

The Canon WB appears to average the data over a central portion of the image. The noise itself should have a mostly random distribution which when averaged results in a consistent value. It's not as accurate as the saturation method but it is easier to implement. What I found is that with certain lighting, it was impossible to saturate all channels. Perhaps the CMOS sensor was implementing some form of anti-blooming logic.

e-k

[_GUI_]

What I find odd, is that if this was the case I would expect the camera to reject the image selected for WB outright as unsuitable.

So the question is was the WB affected at all when these saturated images were selected on the 5D. If so, then the camera accepted the image as a WB image. If the camera accepted the image, then I would hypothesize that the issue could be that all channels weren't fully saturated.

he told me they were, but never all pixels get saturated (think of black pixels and hot pixels). The true is that the 5D guy who tried it on his amera obtained a crazy 15. something multiplier for the B channel. So you can imagine how blue his pictures looked like in the camera's display. It did not reject the file, simply used the few (probably dark) pixels to calculate a wrong WB. But I am just guessing.

The Canon WB appears to average the data over a central portion of the image. The noise itself should have a mostly random distribution which when averaged results in a consistent value. It's not as accurate as the saturation method but it is easier to implement.

I agree.

[tdodd]

I'm trying to work out a foolproof way of getting a perfect set of multipliers by editing the EXIF data directly to force 1024, 1024, 1024, 1024 using ExifTool from here - http://www.sno.phy.queensu.ca/~phil/exiftool/.

At the moment I am unable to fathom how to accomplish this task. It all seems immensely complicated and I can't locate the correct tag name to update. If anyone has a better idea as to how to directly manipulate the multipliers then please reply back. Ditto if you crack the solution using ExifTool before I do - I may never manage it.

[tdodd]

I may have cracked it. Here's the command line I used on my Blown file....


D:\Pictures\Originals\_Imported>exiftool -WB_RGGBLevelsAsShot="1024 1024 1024 1024" test.cr2


I copied the exiftool.exe executable into the directory with my image file, named test.cr2 and ran the above command in a cmd.exe window within the directory. The file got updated exactly as required (I think).

Here are before and after values in the EXIF data. Does this appear to be what is required? If so, this command could be applied to any (EXIF) image file, I think.

Are the "As Shot" values the only ones we need to care about?

[E-K]

I'm trying to work out a foolproof way of getting a perfect set of multipliers by editing the EXIF data directly to force 1024, 1024, 1024, 1024 using ExifTool from here - http://www.sno.phy.queensu.ca/~phil/exiftool/ (http://www.sno.phy.queensu.ca/%7Ephil/exiftool/).

At the moment I am unable to fathom how to accomplish this task. It all seems immensely complicated and I can't locate the correct tag name to update. If anyone has a better idea as to how to directly manipulate the multipliers then please reply back. Ditto if you crack the solution using ExifTool before I do - I may never manage it.

Hi tdodd,

I went down that road initially. The problem is that these tags will have no direct impact. The CWB will utilise the actual image data.

In theory, with a JPEG at least, you could play around with the WB tags to try and cancel things out. I'm not sure about the RAW file as it may not even bother with the WB tags.

When using a JPEG as the CWB image, since you can use any WB on the original image, the camera has to calculate the difference between what the WB should be and what was used as WB in the original image.

I did try using a gray JPEG with some of the WB tags set to (1024 1024 1024 1024) if I remember correctly. Didn't seem to work to well, but I may have just set the wrong tag or missed one.

Maybe I'll give it another go tonight.

e-k

[tdodd]

I tried shooting a completely different image and applied the command above and sure enough, the "As Shot" field updates exactly as expected. But shooting with that image as the custom WB target did not work at all. The resulting image looked pretty normal rather than green tinted. I think there must be a different field that needs to be fixed.

EDIT : I tried updating the "Measured" field as well - no difference :(

[E-K]

I tried shooting a completely different image and applied the command above and sure enough, the "As Shot" field updates exactly as expected. But shooting with that image as the custom WB target did not work at all. The resulting image looked pretty normal rather than green tinted. I think there must be a different field that needs to be fixed.

EDIT : I tried updating the "Measured" field as well - no difference :(

Well evidence points to image data not being a factor so I stand corrected ;). It appears that the custom WB value is precalculated for each image - raw or JPEG - and stored in the image as some metadata. I'll post my results in the Easier UniWB thread. Unfortunately the relevant tag (0xaa) is not writable by ExifTool.

e-k

[tdodd]

I thought I'd mention that I dropped a note to Chuck Westfall to request a built in Uni-WB setting for future cameras and some Uni-WB image files for current bodies and I just received a reply.

Here's my message....

Dear Chuck,
I am very keen that we Canon shooters can access an accurate RGB histogram within our cameras that truly reflects the saturation levels of the sensor, rather than the modified values that are derived from the jpeg preview with its attendant WB adjustment and other processing. I have become aware of the "Uni-WB" custom white balance that is available for Nikon shooters and would love to have the same facility within the Canon camp. I have been working with others to try to create our own version of a Uni-WB calibration image file but it is hard work and not 100% perfect. There are discussion threads on this topic here - http://photography-on-the.net/forum/showthread.php?t=485349 and here - http://photography-on-the.net/forum/showthread.php?t=470141.

I don't think we are asking much here. For the future it would be great to have Uni-WB as a standard choice in the WB settings or even a custom function. Until that time, it would be nice if Canon could construct a Uni-WB calibration file for each camera body.

Would this be possible?

Thanks,
Tim Dodd

and here's his brief, but positive, reply....

Hi, Tim:

I will forward your request to Canon Inc. R&D.

Best Regards,

Chuck Westfall
Technical Advisor/Professional Products Marketing Division
Consumer Imaging Group/Canon U.S.A., Inc.

I don't know whether it would be worth others here adding support for my request, if you have not already done so, but, if you want to add your backing, I wrote to Chuck at this address....

cwestfall@cusa.canon.com

[mangaloreaviators]

Hello All,

In order to take the FAQ's on Histogram thread more usefull and help members master the Histogram I suggest the following things.

1. A set of 50 Questionairs which has 4 answers as option. Eash question will carry 2 points which makes a score total of 100Points.
Need a help of Software Engg who can create a Webpage which will automatically calculate the points by cross checking answers.

2. Questions should start from easy and end with difficult ones.

3. A Guide prepared under POTN Label is also welcome so that members can read them before taking the review

4. 5 Match the following questions- Side A will have Images and Side B will have corresponding Histograms- member has to choose the right histogram for the image.
***This will be for Brightness Histogram***

5. 5 Match the following questions- Side A will have Images and Side B will have corresponding Histograms- member has to choose the right histogram for the image.
***This will be for RGB Histogram***

Finally who complete the review with more than 95% Score will get a POTN Gold Medal or E- Certificate for passing the Histogram test. Just a suggestion.
Let me know your thoughts. In the mean time I will also check with some of my friend who work for Software firm to design something like this.

Moderators- any thoughts?

[tdodd]

To be honest that seems an awful lot of effort and I really can't see what it would gain. How on earth are you going to assemble 50 images with meaningfully different histograms? Maybe I'm a genius or maybe I'm so stupid that I'm completely missing the point but I don't get what is so hard about this....

There are four basic histogram scenarios that are important....

- histogram data fits completely within the edges of the histogram, with no spikes on the extreme left or right - you have successfully captured the entire tonal range of the scene;

- histogram data has spikes on the extreme left and right edge - the scene contains such a wide dynamic range that the camera can not record both the extremes of darkness and brightness at the same time. You need to decide whether the image is OK as it is, or if you need to increase the exposure a little to include more shadow detail, or reduce it a little to include more highlight detail. If you can add light to the darker areas then maybe you can reduce the dynamic range of the scene and have everything fit;

- histogram has a spike on the extreme left, but not on the right - you have lost detail in the dark areas, where some parts of the image are pure, featureless black. If there is space on the right hand side you can increase your exposure to shift the histogram to the right and reveal more shadow detail;. Even if there is not space on the right you can still choose to increase the exposure to reveal more shadow detail, but you will start to lose highlight detail if you start creating a spike on the right hand edge;

- histogram has a spike on the extreme right, but not on the left- you have lost detail in the bright areas, where some parts of the image are pure, featureless white. If there is space on the left hand side you can reduce your exposure to shift the histogram to the left and reveal more highlight detail. Even if there is not space on the left you can still choose to reduce the exposure to reveal more highlight detail, but you will start to lose shadow detail if you start creating a spike on the left hand edge;


Sure, the shape of the histogram between the extreme left and extreme right will vary all over the place, from one scene to the next but all those variations don't really matter. In terms of fully understanding histograms you can see how the bumps relate to the scene, but the tutorials on those things are already on the first page of this thread. We really do not need to plough through 50 examples, IMHO. The important thing is to understand what is going on at or near the edges.

[mangaloreaviators]

tdodd- I'm working on this and get back shortly.

Sorry there was a delay in responding as I was busy with Interstate Rowing Competetion that took place.

[adas]

uni-what?
nikon lust much?
You people should keep in mind that when nikon can't keep the attention on themselves with true technological advances, they try with exotic overnight features.

[TJA]

This is a very interesting topic ...
Sadly, it seems to be dead?

I also am hunting a good uniWB and ways how to get one - sadly, Nikon does not allow to use the "complete blown" or "completely dark" shots for WB ...