Help me understand why highlights can appear blown but are recoverable

[thekev]

Originally Posted by VirtualRain

To help better illustrate my point, I believe this is the situation... as I've illustrated below. The Normal Histogram we see is misleading in that it implies data is clipped when it is not. I guess I'm wondering why we can't have a proper RAW histogram that actually shows when data is really being clipped? Is this technically impossible for some reason or are camera and software companies lazy, or am I the only one that sees a need for this?

Technically lazy isn't exactly the case. There are multiple aspects to it, and if anything marketing departments are the lazy ones. They try to make things too digestible at times, which results in a lot of inaccurate intuition.

Originally Posted by paolo-

I by no means know the answer and am by no means competent in the matter. I think it's because of the way the RAW files are made. Say current Canon cameras use 14bit raw files. However, a camera like a 7d spreads it's dynamic range over something like 12.x bits. (not sure where I read that *citation needed) So if the RAW converter showed you the full 14 bits you could have a clipped image that doesn't look clipped on the histogram. Also the dynamic range of the camera isn't set in stone, you know how you can still get data from an over exposed image but it's ugly and noisy, where do you draw that line? Also that lines tends to move with the ISO at which the image was shot and probably temperature (not to mention between same model cameras).

To add to this, the number of bits stored is independent of the range described. It's just how many values are used to describe that range. It can get a bit confusing at times when you look at things like 32 bit floating point modes in different software where they don't just reduce rounding errors but provide a way to store values over a greater range. What often confuses people is that this is a function of the way the format is set up rather than the number of bits that describe it. They just need a lot of bits to make such a descriptive method work without banding. In the case of digital cameras, you're looking at camera data with a dynamics range that exceeds typical working profiles and greatly exceeds display gamuts. The raw data isn't gamma corrected. It's not necessarily precisely linear in its nature, but it's not baked/profiled to a specific gamma encoding. Anyway a lot of cameras have a greater range than typical working profiles, so sometimes you have leverage. If they're really blown out, sometimes 1-2 channels may still contain detail. De-bayering data has some interpolation involved either way. In that case it could be way more, thus the reduced quality. Blah I wish I could better organize this.

Originally Posted by ChrisA

The problem is more complex and there are a few dfferent cases

1) The JPG file is "clipped" and has some pure white areas while the raw file is not clipped

It's partly that they have different clipping points. You'll sometimes see it referred to as clamped data as well.

Originally Posted by VirtualRain

Agreed, but it doesn't explain why we're only given a JPEG histogram to work with in any and all tools.

I'm not sure. It could be they want you to see the results as they should normally look. It's typically more extreme lighting situations where you would need to pull out a lot of extra range at the edges, although raw processing has some tools for dealing with things such as extremely saturated details where the resulting color values would otherwise cause clipping in some channels with default processing and relative colorimetric conversion.

Originally Posted by Edge100

The histogram represents the in-camera JPEG conversion - which has a contrast curve applied to it, not the Raw file - which has linear gamma.

The camera is "squashing" your highlights into a JPEG.

Well gamma encoding isn't necessarily evil. Without that how would you really view things meant to appear continuous in tone over a limited dynamic range?

Originally Posted by VirtualRain

I suspect you're right, but we should start demanding better tools. Wouldn't it be helpful to have the choice of histograms presented in post #3? So you could tell what is clipped in the JPEG, what is recoverable from RAW data, and what is clipped in RAW?

The tools have evolved somewhat relative to computing power. Ideally you'd be able to access that entire range while a program like photoshop merely interprets it to display as gamma 2.2 Adobe RGB or sRGB or whatever else you choose (not opening that debate). It's not like your display can handle the number of values that would be needed to properly show something without heavy gamma correction. The fact that the most popular raw processors haven't come up with something like an export to 32 bit .EXR or .HDR probably means it's a niche request.

[Prodo123]

Originally Posted by VirtualRain

To help better illustrate my point, I believe this is the situation... as I've illustrated below. The Normal Histogram we see is misleading in that it implies data is clipped when it is not. I guess I'm wondering why we can't have a proper RAW histogram that actually shows when data is really being clipped? Is this technically impossible for some reason or are camera and software companies lazy, or am I the only one that sees a need for this?

You can access this "expanded histogram" albeit not directly in the histogram itself!
The default histogram is a representation of the RAW file in a 24-bit RGB color space, which means all the colors must fit within (255,255,255). Anything above that is clipped and is beyond the 24-bit color space and cannot be represented by that system. Since the final image is to be in 24-bit also, the histogram shows how the RAW file conforms to the 24-bit color space.

You can see the RAW histogram which you pointed out in the expanded curves plot, which overlays the histogram over the curve for a visual representation of the curve edits. This can be done in Aperture, not sure about other editing programs.

[VirtualRain]

Originally Posted by Prodo123

You can see the RAW histogram which you pointed out in the expanded curves plot, which overlays the histogram over the curve for a visual representation of the curve edits. This can be done in Aperture, not sure about other editing programs.

Interesting... I need to explore this... here's the elephant's RAW histogram in Aperture using the extended curve range...

[kallisti]

Isn't the answer to your question simply that DSLRs don't display accurate data in their histograms? Perhaps there is a technical reason they don't do so (though I question this given the ability of Leica to do it in a monochrome sensor). Or it's just camera makers are lazy with their features, assuming that no one will notice or care. We should all bitch so the camera makers include a feature that is actually useful, like histograms that actually reflect what the sensors are seeing

[Laird Knox]

Originally Posted by kallisti

Isn't the answer to your question simply that DSLRs don't display accurate data in their histograms? Perhaps there is a technical reason they don't do so (though I question this given the ability of Leica to do it in a monochrome sensor). Or it's just camera makers are lazy with their features, assuming that no one will notice or care. We should all bitch so the camera makers include a feature that is actually useful, like histograms that actually reflect what the sensors are seeing

Unfortunately you can't fit 14 bits into 8.

[kevinfulton.ca]

Originally Posted by VirtualRain

I suspect you're right, but we should start demanding better tools. Wouldn't it be helpful to have the choice of histograms presented in post #3? So you could tell what is clipped in the JPEG, what is recoverable from RAW data, and what is clipped in RAW?

I get what you're saying, but honestly not in the field. It's best to avoid "recovering" over exposed images even with RAW. Just because it can be recovered doesn't mean it will look good. All your color info will be different then a properly exposed image.Think of the JPEG histogram as a "safe zone". If you stay inside that, minus some off subject specular highlights or shadows, you'll have a lot more room once those images get to post.

[kallisti]

Originally Posted by Laird Knox

Unfortunately you can't fit 14 bits into 8.

Forgive my ignorance here, but wouldn't it be possible to have the individual histograms representing each color channel actually display the RAW data available for that channel? Not talking about a combined histogram.

I got burned several years ago while shooting pink/red flowers in Hawaii on a Nikon body. The meter showed an adequate exposure. The histogram showed an adequate exposure (this was on a body that only displayed a white "overall" histogram--i.e. the green channel). The actual image was completely blown out with no details present. I learned my lesson and started viewing the individual color channel histograms once I got a body that could display that info.

The ability to view individual color channels on the histogram is a huge improvement. The fact that they are based on an in-camera JPG and not the actual RAW data is...strange.

I agree with the idea that the histogram should serve as a "safe zone." However, I think it is more useful to have a true sense of what is present in the RAW data as this can impact exposure decisions at the time of capture.

[kevinfulton.ca]

Originally Posted by kallisti

Forgive my ignorance here, but wouldn't it be possible to have the individual histograms representing each color channel actually display the RAW data available for that channel? Not talking about a combined histogram.

I got burned several years ago while shooting pink/red flowers in Hawaii. The meter showed an adequate exposure. The histogram showed an adequate exposure (this was on a body that only displayed a white "overall" histogram--i.e. the green channel). The actual image was completely blown out with no details present. I learned my lesson and started viewing the individual color channel histograms once I got a body that could display that info.

The ability to view individual color channels on the histogram is a huge improvement. The fact that they are based on an in-camera JPG and not the actual RAW data introduces a limit to their usefulness.

I agree with the idea that the histogram should serve as a "safe zone." But it's also annoying that you can't tell based on the histograms how much wiggle room you really have when shooting.

While you're correct that the green channel does contain the most amount of detail it is not what is used for the combined channels. What you are seeing are the combined values of all three colors in the lows (shadows), mids, and highs (highlights) of the image. If it was a red flower then the green channel on your histogram would have only shown barely any information on the screen accept for a tiny bit of info in the lows (shadows) If your image was blown out then there was either something funky going on with your metering or a change in light may have happened once you were already in manual mode.

If you're in a high contrast situation then it's always best to plan ahead (if possible) and shoot some bracket exposures. The other to do is ask yourself "will this under, or over, exposed section of the photo take away from the subject". If the sun and specular highlights are clipping, but my subject is properly exposed does it really take away from the photo? Most of time the answer will be "no". This will always be a choice that you'll have to make not matter how good your histogram is.

[kallisti]

Originally Posted by kevinfulton.ca

While you're correct that the green channel does contain the most amount of detail it is not what is used for the combined channels. What you are seeing are the combined values of all three colors in the lows (shadows), mids, and highs (highlights) of the image.

I stand somewhat corrected. The combined histogram on my D800 is "sort of" a combination of all the channels. However, it is seriously weighted against certain channels.

As an example, here is an image of the D800 display from an image of the red cover of my iPad.



Note that even though the combined white channel shows some highlight clipping ("blinkies" on upper left only), the red channel is completely blown out, far out of proportion to what the "combined" channel would suggest (if you look for "blinkies" when viewing the red channel, the entire cover is blown out). Since the subject is red, there is actually no usable detail here and the image is horribly over exposed. The meter was fooled and if you were basing exposure on the combined histogram you wouldn't have known this.

[kevinfulton.ca]

Originally Posted by kallisti

[The content of this post was incorrect]

You got me questioning myself. Luckily I had my camera next to me for testing. The shape looks similar, but you can see that they are in fact different. Maybe it's just my camera, but if my old Canon XTi has it and your Nikon D800 doesn't then yikes! Somebody call Nikon quick!

[Laird Knox]

Originally Posted by kallisti

Forgive my ignorance here, but wouldn't it be possible to have the individual histograms representing each color channel actually display the RAW data available for that channel? Not talking about a combined histogram.

I got burned several years ago while shooting pink/red flowers in Hawaii on a Nikon body. The meter showed an adequate exposure. The histogram showed an adequate exposure (this was on a body that only displayed a white "overall" histogram--i.e. the green channel). The actual image was completely blown out with no details present. I learned my lesson and started viewing the individual color channel histograms once I got a body that could display that info.

The ability to view individual color channels on the histogram is a huge improvement. The fact that they are based on an in-camera JPG and not the actual RAW data is...strange.

I agree with the idea that the histogram should serve as a "safe zone." However, I think it is more useful to have a true sense of what is present in the RAW data as this can impact exposure decisions at the time of capture.

When I reread the post I realized my 14 into 8 comment wasn't entirely accurate. On further thought the histogram is a representation of the data so it could "represent" either 8 or 14-bit data.

While there will be some difference between a RAW and JPEG histogram of the same image they should be pretty close to identical. An 8-bit red of 255 will be darn close to a 14-bit value of 16380.

As for the difference in the channels, the "white" histogram is a mapping of tonal values. So that peak you see is what happens when you add up the red, green and blue channels. The individual channels might show crazy variances like your example but the actual tones when combined result in a curve that spans the middle of the graph.

The auto exposure in modern cameras is pretty darn amazing but it still doesn't match millions of years of evolution. In certain cases the camera will think it is doing a good job but one particular channel might be blow out. Sometimes the only way to get around this is to bracket the shot with multiple exposures.

Overall this has been a good thread and has had me thinking through some of the details of what is going on. This is a good thing as the more you understand the process the better you will be able to get exactly what you want.

[kallisti]

Originally Posted by kevinfulton.ca

You got me questioning myself. Luckily I had my camera next to me for testing. The shape looks similar, but you can see that they are in fact different. Maybe it's just my camera, but if my old Canon XTi has it and your Nikon D800 doesn't then yikes! Somebody call Nikon quick!

See my corrected post above. I pulled out my D800 and shot a red subject. The camera metering seems to reflect the combined channel which is weighted heavily towards the green channel. In most shooting, this isn't an issue. When shooting red subjects, there is a very real chance for overexposure.

This is true for every Nikon digital body I have owned. Can't speak for Canon (though from what I've read it applies there as well). Shoot a red subject without other colors present and see what your histograms show.

At least on Nikon bodies, the combined histogram isn't an accurate reflection (i.e. a summation) of the data from each channel.

(1) it is based on an in-camera JPG rather than RAW sensor data
(2) it is biased towards the green channel

This has implications when basing exposure on the image histogram (the "combined" channel can be misleading, especially if shooting a largely red subject) and the individual channels may not accurately reflect usable data that the sensor is actually seeing.

[ChrisA]

Originally Posted by VirtualRain

...
So, a RAW histogram does exist, it is useful, and it would be great if cameras implemented this (IMHO).

I can't figure out how to prove it but I'm pretty sure my Nikon D200 has a raw histogram because it can display four histograms, one for each channel and one for white. It does this even when I ask the camera not to produce a JPG file. When the camera says the pixel is clipped it really is clipped when I look later on the computer. It will also blink the blown red, green or blue pixels or I can have it blink the blown white ones. It appears to be working at the raw level.

I don't know about the newer or the lower end Nikon SLRs but mine does what you are asking for.

----------

Originally Posted by kallisti

Forgive my ignorance here, but wouldn't it be possible to have the individual histograms representing each color channel actually display the RAW data available for that channel? Not talking about a combined histogram.

I got burned several years ago while shooting pink/red flowers in Hawaii on a Nikon body. ....

Which Nikon body? My Nikon D200 has this feature and it works just like what you are asking for.

I used to have a Nikon D50 and it only had the white histogram. I guess when you buy a camera this is one more thing to look for

[kallisti]

Originally Posted by ChrisA

I can't figure out how to prove it but I'm pretty sure my Nikon D200 has a raw histogram because it can display four histograms, one for each channel and one for white. It does this even when I ask the camera not to produce a JPG file. When the camera says the pixel is clipped it really is clipped when I look later on the computer. It will also blink the blown red, green or blue pixels or I can have it blink the blown white ones. It appears to be working at the raw level.

You are incorrect. The histogram is being displayed based on an in-camera JPG regardless of which settings you have made on your D200. The only camera body that displays a histogram based on the actual RAW data is the Leica M Monochrome.

Originally Posted by ChrisA

Which Nikon body? My Nikon D200 has this feature and it works just like what you are asking for.

I used to have a Nikon D50 and it only had the white histogram. I guess when you buy a camera this is one more thing to look for

I may have misspoken. After going back through my photo library, it was a Nikon D300 and I can't remember if it had the ability to look at individual color channels or not. If it did, at the time I was inexperienced and assumed that the combined "white" channel was truly reflective of the overall exposure. It wasn't and I got burned. On a series of shots of pink/red flowers, the camera meter showed correct exposure and the white histogram didn't show highlight clipping. The resulting images were horribly over exposed and not usable. Lesson learned. When shooting red subjects on a Nikon, don't trust the in-camera meter.

[kevinfulton.ca]

Originally Posted by kallisti

I stand somewhat corrected. The combined histogram on my D800 is "sort of" a combination of all the channels. However, it is seriously weighted against certain channels.

As an example, here is an image of the D800 display from an image of the red cover of my iPad.

Image

Note that even though the combined white channel shows some highlight clipping ("blinkies" on upper left only), the red channel is completely blown out, far out of proportion to what the "combined" channel would suggest (if you look for "blinkies" when viewing the red channel, the entire cover is blown out). Since the subject is red, there is actually no usable detail here and the image is horribly over exposed. The meter was fooled and if you were basing exposure on the combined histogram you wouldn't have known this.

I'm starting to understand the issue you were having with that flower. I don't think you quite know how to read a histogram. I'm not trying to be a jerk I'm just trying to help so please hear me out before anybody gets heated.

Levels on the left= Shadows, middle= mids, and right= highlights. The height of the levels shows you the amount of information in that area. If the highlight levels are pushed up against the left side with no gap that is showing you that some highlights are blown out (in the case a good chunk). Pushed against the right side and that shadows will be crushed (in this case there's a gap there so you're good). The height of the level shows how much detail there is at any given part of the image (the higher the peak the more detail). So lets break down this photo of your cover.

If I'm looking at the combined histogram I can see that there's a good chunk of highlights (right side) that are slammed against the edge. So at that point I know you're over exposed and you'll need to compensate by a couple of stops. At this point you could look at the red channel and see that it's mostly red highlights that are clipping, but that doesn't really matter because we already know from the combined histogram that it's clipping and needs to be adjusted. Where the separate histogram a really come into play is for white balance. If you shoot a grey card they should all be aligned almost perfectly if you have the correct white balance. If your grey card is all over the place then you know that you need to fine tune it.

Again, not trying to be a jerk. Just trying to help out best I can. Hope this clarifies things.

Taking a good look at things my "combined" histogram is listed on my camera as "brightness" so this is probably only showing the luminance of the image and not all the colour details. Chances are it's the same with your Nikon.

[kallisti]

Originally Posted by kevinfulton.ca

I'm starting to understand the issue you were having with that flower. I don't think you quite know how to read a histogram. I'm not trying to be a jerk I'm just trying to help so please hear me out before anybody gets heated.

Levels on the left= Shadows, middle= mids, and right= highlights. The height of the levels shows you the amount of information in that area. If the highlight levels are pushed up against the left side with no gap that is showing you that some highlights are blown out (in the case a good chunk). Pushed against the right side and that shadows will be crushed (in this case there's a gap there so you're good). The height of the level shows how much detail there is at any given part of the image (the higher the peak the more detail). So lets break down this photo of your cover.

If I'm looking at the combined histogram I can see that there's a good chunk of highlights (right side) that are slammed against the edge. So at that point I know you're over exposed and you'll need to compensate by a couple of stops. At this point you could look at the red channel and see that it's mostly red highlights that are clipping, but that doesn't really matter because we already know from the combined histogram that it's clipping and needs to be adjusted. Where the separate histogram a really come into play is for white balance. If you shoot a grey card they should all be aligned almost perfectly if you have the correct white balance. If your grey card is all over the place then you know that you need to fine tune it.

Again, not trying to be a jerk. Just trying to help out best I can. Hope this clarifies things.

Taking a good look at things my "combined" histogram is listed on my camera as "brightness" so this is probably only showing the luminance of the image and not all the colour details. Chances are it's the same with your Nikon.

I'm not sure how to best respond. This discussion isn't related to the proper exposure of a given digital image. I understand how to read a histogram, though I'm a bit confused by some of your comments relating to this (you switched left and right in there--blown highlights are to the right and lost shadows are to the left). One of my points is that the combined histogram isn't valid for some images, the one I posted which prompted this digression being a case-in-point. I'm not sure how white point entered into the discussion at all.

The argument relates to the algorithms involved in how digital cameras determine proper exposure via their built-in meters (and display that information in histograms which are intended to help the photographer at the time of image capture).

It's an argument about technical issues relating to how camera hardware/software actually works which may not jive with how we assume it should work. Not an argument about the basics of photographic technique/post-processing.

To be blunt: Nikon camera bodies (and possibly bodies from other manufacturers) completely blow when metering red subjects. They tend to over-expose in these circumstances. I invoked an anecdotal example from years ago with a D300 body. I provided an actual example off-the-cuff with my D800 showing this problem still exists. Red subjects confuse the metering system on Nikon bodies (and possibly Canons?). What the histograms I provided are showing is that the camera thinks it is giving a "good" exposure while in reality it is completely screwing up when faced with a red subject The result is horrible over-exposure. This is kind of a reverse example of what the OP started this thread about--histograms don't display accurate information.

[kevinfulton.ca]

Originally Posted by kallisti

I'm not sure how to best respond. This discussion isn't related to the proper exposure of a digital image. I understand how to read a histogram or set a proper white point.

The argument relates to the algorithms involved in how digital cameras determine proper exposure (and display that information in histograms which are intended to help the photographer at the time of image capture).

It's an argument about technical issues relating to image capture and not the fundamentals of photography.

To be blunt: Nikon camera bodies (and possibly bodies from other manufacturers) completely blow when metering red subjects. They tend to over-expose in these circumstances. I invoked an anecdotal example from years ago with a D300 body. I provided an actual example off-the-cuff with my D800 showing this problem still exists. Red subjects confuse the metering system on Nikon bodies (and possibly Canons?). What the histograms I provided are showing is that the camera thinks it is giving a "good" exposure while in reality it is completely screwing up when faced with a red subject The result is horrible over-exposure.

I get that we're getting off topic. Maybe Nikon's do have trouble with reds, but my point is that the histogram you posted shows is blown out in the standard histogram. Seeing the red histogram is not necessary. The standard histogram already shows it. Perhaps it's the clipping flashes that are off, but the standard histogram seems to be accurate.

[VirtualRain]

Originally Posted by kallisti

I stand somewhat corrected. The combined histogram on my D800 is "sort of" a combination of all the channels. However, it is seriously weighted against certain channels.

As an example, here is an image of the D800 display from an image of the red cover of my iPad.



Note that even though the combined white channel shows some highlight clipping ("blinkies" on upper left only), the red channel is completely blown out, far out of proportion to what the "combined" channel would suggest (if you look for "blinkies" when viewing the red channel, the entire cover is blown out). Since the subject is red, there is actually no usable detail here and the image is horribly over exposed. The meter was fooled and if you were basing exposure on the combined histogram you wouldn't have known this.

Very interesting and illustrative example IMHO.

So to re-iterate, my point is that although the red channel appears completely blown out in this histogram, the fact is, it was probably not actually clipped by the sensor. Something we can't possibly know without looking at a histogram of the RAW data.

It would be nice to know by looking at a RAW histogram if the red channel really was clipped by the sensor so you could retake the photo to eliminate clipping, or just leave it and recover it with a modified curve in post if it was not clipped.

Without a RAW histogram, all you can do is be safe and retake the photo based on this histogram, lowering the exposure to ensure the red channel is not blown but in doing so, under expose the other channels, possibly introducing unnecessary noise in the process. Not ideal... especially if you're trying to address a problem that's not there.

I don't think anyone can argue that having the option of a RAW histogram is a bad thing. It may not be helpful in all cases, but in this case, and in any high dynamic range situation, it would be great to be armed with as much information about your exposure as possible.

[kallisti]

Originally Posted by kevinfulton.ca

I get that we're getting off topic. Maybe Nikon's do have trouble with reds, but my point is that the histogram you posted shows is blown out in the standard histogram. Seeing the red histogram is not necessary. The standard histogram already shows it. Perhaps it's the clipping flashes that are off, but the standard histogram seems to be accurate.

The standard histogram is *not* accurate. According to the standard histogram, the only area that is blown out is the upper left of the image. According to the standard histogram, the iPad cover is properly exposed. The iPad cover is actually overwhelmingly over-exposed with zero detail present in the image. But the camera metering system was blind to this at the time of capture. Short of making a movie which documents what the camera considered clipped for each channel, I can't show this to you. Please just trust me. Yes, the overall image was slightly over-exposed according the camera via the standard histogram. The camera thought this only related to the upper left corner. The camera was *wrong*, horribly wrong. Yes it's an awesome Nikon that costs a huge amount of money. But it was *wrong* in metering this image, completely overwhelmingly *wrong* in every way. Had the same standard histogram been there for a non-red subject I could have salvaged this hypothetical image. But because it is a red image, there is nothing to salvage as the red channel is blown and it is the only channel that matters for this subject.

I'm repeating myself. And I'm kicking myself for getting drawn into this exchange. I've said what I wanted to say. Hopefully some will find it helpful.

[kevinfulton.ca]

Originally Posted by kallisti

The standard histogram is *not* accurate. According to the standard histogram, the only area that is blown out is the upper left of the image. According to the standard histogram, the iPad cover is properly exposed. The iPad cover is actually overwhelmingly over-exposed with zero detail present in the image. But the camera metering system was blind to this at the time of capture. Short of making a movie which documents what the camera considered clipped for each channel, I can't show this to you. Please just trust me. Yes, the overall image was slightly over-exposed according the camera via the standard histogram. The camera thought this only related to the upper left corner. The camera was *wrong*, horribly wrong. Yes it's an awesome Nikon that costs a huge amount of money. But it was *wrong* in metering this image, completely overwhelmingly *wrong* in every way. Had the same standard histogram been there for a non-red subject I could have salvaged this hypothetical image. But because it is a red image, there is nothing to salvage as the red channel is blown and it is the only channel that matters for this subject.

I'm repeating myself. And I'm kicking myself for getting drawn into this exchange. I've said what I wanted to say. Hopefully some will find it helpful.

Ok ok ok let's all just take a deep breath here. There's obviously a miscommunication and if I offended you or frustrated you I am deeply sorry. Never my intent. Just trying to help here by offering a POSSIBLE solution to, what seems to be, a constant source of frustration. That being said communication can get a little out of whack on here for obvious reasons. I just want to clarify that we're talking about the same thing and if I completely missed the point then I'll apologize even more. So just bare with me here.

Here's the picture of your cameras histogram (or maybe it's not and that's one of the things I missed). The red arrow is pointing to a peak that shows all the detail in the mid tones of the standard histogram, which are NOT clipping. The green is pointing to the highlight levels on the standard histogram which ARE clipping and will cause the over exposure you're talking about (you can see how they're pushed against the right wall). Are you suggesting that it should be read the other way around? (This is what I thought might have been going on)

OR were you saying that, even though the standard histogram is showing that there is a good amount of clipping that it's far worse then displayed because of Nikon's red issues? If it's the second one then I apologize BIG TIME (over top of my existing apology ), because I was just not picking that up. It's probably because it may be more of a Nikon thing so I never experienced or heard of such an issue. For the record I wasn't putting anybody down for having a super expensive body that has said issue (I'd love to have one myself), but it really does suck that Nikon hasn't addressed it. Hopefully a firmware update can eventually fix it if that's what's happening.

Trust me that I'm not a snotty guy trying to rope anybody into anything because he's got nothing better to do. This is probably just a case of misunderstanding. Nothing more. I'm trying to help and or learn, not infuriate.

[ChrisA]

Originally Posted by kallisti

I stand somewhat corrected. The combined histogram on my D800 is "sort of" a combination of all the channels. However, it is seriously weighted against certain channels.
....

I think My D200 works like your D800. The display is slightly different but not by much. To make the white display, what I think they do is add all three channels and then only blink the pixels that are just below the maximum value.

THis is still a pretty good exposure tool once you learn to read it. The RGB histogram is very useful. But then I used to be able to shoot decent exposures based on just reading the inside of the box that the film came it. It said in bright sunlight to shoot at 1/(film speed) at f/16 and that works perfectly and then it said to open up one stop for clouds. Those instructions still work with dSLRs.

[kallisti]

Okay, will attempt to address both issues that are running through this thread.

(1) issue raised in the OP--why are highlights appearing blown out on the histogram but actually have detail when viewing the RAW data of the image.
(2) the standard histogram isn't displaying accurate information regarding exposure, specifically relating to the red channel.



Note that the common channel reflects an adequate exposure with only a little bit blown out on the right. The red channel shows that quite a bit more of the image is actually blown out and since the subject is red (well, orange in this case), the red channel is actually over-exposed. For this color image it isn't a major problem, but if you were going to convert this to B&W and use the red channel for your source channel, you would notice the over-exposure. Note well that all of these comments are based on the histogram displayed by the camera (which is based on a JPG). It does *not* relate to the actual data captured by the sensor.

Here is the histogram displayed in Aperture based on the RAW data from the camera:


Note that the red channel isn't actually blown out and there is usable information there. Note further that there really isn't significant clipping at all. Had I captured this image as a JPG, then all of my above comments would be valid. Because I shot in RAW, none of them actually apply. You could use the red channel for a B&W conversion--there is usable detail there.

I will state again that the camera histogram is based on a JPG and *not* the RAW data. So the information displayed on the in-camera histogram does not reflect what the sensor actually captured. Also, the histogram is biased towards the green channel, which in some circumstances can cause problems when shooting red (or possibly even blue) subjects.

In this example, the camera actually obtained correct exposure even though the individual histograms might not have shown this. There are other cases in my experience where the camera actually missed exposure and clipped the reds even in the RAW data. Bottom line--you can't trust the histogram on the camera. This may also serve as an example of why it is better to shoot in RAW.

[kevinfulton.ca]

Originally Posted by kallisti

Okay, will attempt to address both issues that are running through this thread.

(1) issue raised in the OP--why are highlights appearing blown out on the histogram but actually have detail when viewing the RAW data of the image.
(2) the standard histogram isn't displaying accurate information regarding exposure, specifically relating to the red channel.

Image

Note that the common channel reflects an adequate exposure with only a little bit blown out on the right. The red channel shows that quite a bit more of the image is actually blown out and since the subject is red (well, orange in this case), the red channel is actually over-exposed. For this color image it isn't a major problem, but if you were going to convert this to B&W and use the red channel for your source channel, you would notice the over-exposure. Note well that all of these comments are based on the histogram displayed by the camera (which is based on a JPG). It does *not* relate to the actual data captured by the sensor.

Here is the histogram displayed in Aperture based on the RAW data from the camera:
Image

Note that the red channel isn't actually blown out and there is usable information there. Note further that there really isn't significant clipping at all. Had I captured this image as a JPG, then all of my above comments would be valid. Because I shot in RAW, none of them actually apply. You could use the red channel for a B&W conversion--there is usable detail there.

I will state again that the camera histogram is based on a JPG and *not* the RAW data. So the information displayed on the in-camera histogram does not reflect what the sensor actually captured. Also, the histogram is biased towards the green channel, which in some circumstances can cause problems when shooting red (or possibly even blue) subjects.

In this example, the camera actually obtained correct exposure even though the histograms might not have shown this. There are other cases in my experience where the camera actually missed exposure and clipped the reds even in the RAW data. Bottom line--you can't trust the histogram on the camera. This may also serve as an example of why it is better to shoot in RAW.

Got ya. So basically I completely misunderstood what you were saying lol! Sorry about that! This illustrates everything very nicely and makes perfect sense to me. I'm guessing the fact that I've never experienced such a thing with my camera played into the misunderstanding a fair bit. Or maybe my camera does do it, but just not to the same degree? Either way I will definitely take that into consideration next time I'm shooting a red subject. Thanks for the info and very sorry again for the misunderstanding!! I'm also from Vancouver so send me a PM if my apology needs beer attached

[kallisti]

Originally Posted by kevinfulton.ca

Got ya. So basically I completely misunderstood what you were saying lol! Sorry about that! This illustrates everything very nicely and makes perfect sense to me. I'm guessing the fact that I've never experienced such a thing with my camera played into the misunderstanding a fair bit. Or maybe my camera does do it, but just not to the same degree? Either way I will definitely take that into consideration next time I'm shooting a red subject. Thanks for the info and very sorry again for the misunderstanding!! I'm also from Vancouver so send me a PM if my apology needs beer attached

You are good, no worries

The issue is a very real issue though. It can be frustrating to take the time to "chimp" and view camera histograms after taking an image with the assumption that the histograms are displaying accurate information that allows you to adjust subsequent exposures based on what you see in the histograms (either the combined channel or individual channels).

It's annoying (to put it mildly) that the histograms aren't displaying the actual information available to the sensor. If you are shooting in JPG you are set, and what you see is what you get. However, there is more information available to the sensors that you capture when shooting RAW. The histograms don't show you this though. Flaw in almost all modern bodies, with the only exception to my knowledge at the time of this post being the Lieca M Monochrom.

[kevinfulton.ca]

Originally Posted by kallisti

You are good, no worries

The issue is a very real issue though. It can be frustrating to take the time to "chimp" and view camera histograms after taking an image with the assumption that the histograms are displaying accurate information that allows you to adjust subsequent exposures based on what you see in the histograms (either the combined channel or individual channels).

It's annoying (to put it mildly) that the histograms aren't displaying the actual information available to the sensor. If you are shooting in JPG you are set, and what you see is what you get. However, there is more information available to the sensors that you capture when shooting RAW. The histograms don't show you this though. Flaw in almost all modern bodies, with the only exception to my knowledge at the time of this post being the Lieca M Monochrom.

I'm wondering if this is something that camera manufacturers will address soon as sensors continue to get more and more sensitive with broader dynamic range. That could be one reason it may not be as noticeable on my camera (it is around 5 years old after all). Its 10MP sensor just doesn't have the dynamic range or data collection of your D800's (or any other newer camera for that matter). Right now they're probably thinking "jpeg readouts are good enough", but do people buying higher end DSLR's shoot JPEG? I will sometimes shoot JPEG+RAW, but I can't remember the last time I shot strictly JPEG.