Histograms - Theory Question

Kent · Dec 16, 2001

As I understand the theory of the histogram, the graph is a representation of the distribution of grey scale from pure black to pure white. Colors are converted to a black and white equivalent.

The vertical height of any point is a measure of the number of pixels (or some other small unit of measure) at that particular point on the horizontal axis (from black at the left to white at the right).

Since evey camera at any specific setting has a fixed number of pixels or units for analysis in generating a histogram, I would expect the area under the curve of a histgram to be constant for any camera. The only exception would be in situations where there is so much of a particular shade in the horizontal axis, that the vertical reading goes off the top of the histogram display (which raraely happens on an image that is reasonably well exposed, unless there subject is mostly a single shade).

On my camera (Dimage 7), the area of the histogram varies considerably from shot to shot. Can anyone explain why?

Kent
San Francisco--Kent

Andrew Grant27981 · Dec 16, 2001

The Y axis scale is probably poportional instead of absolute.

Kent said:
As I understand the theory of the histogram, the graph is a
representation of the distribution of grey scale from pure black to
pure white. Colors are converted to a black and white equivalent.

The vertical height of any point is a measure of the number of
pixels (or some other small unit of measure) at that particular
point on the horizontal axis (from black at the left to white at
the right).

Since evey camera at any specific setting has a fixed number of
pixels or units for analysis in generating a histogram, I would
expect the area under the curve of a histgram to be constant for
any camera. The only exception would be in situations where there
is so much of a particular shade in the horizontal axis, that the
vertical reading goes off the top of the histogram display (which
raraely happens on an image that is reasonably well exposed, unless
there subject is mostly a single shade).

On my camera (Dimage 7), the area of the histogram varies
considerably from shot to shot. Can anyone explain why?

Kent
San Francisco
--
Kent

john g · Dec 16, 2001

I think that you're mixing up terms, and the historgram can be understood easily for practical work, much less so for speculation. I know how to use it in Photoshop, and I assume you know how to use it practically for your applications, so perhaps you want to rephrase?

Kent said:
As I understand the theory of the histogram, the graph is a
representation of the distribution of grey scale from pure black to
pure white. Colors are converted to a black and white equivalent.

Well, not really a representation of grey scale, it shows all the colors in a term you might call: brightness, luminosity, value, density, or % of total ink.

Member said:
The vertical height of any point is a measure of the number of
pixels (or some other small unit of measure) at that particular
point on the horizontal axis (from black at the left to white at
the right).

Not pixels at all, some other small unit.

Member said:
Since evey camera at any specific setting has a fixed number of
pixels or units for analysis in generating a histogram, I would
expect the area under the curve of a histgram to be constant for
any camera.

Not fixed, but not infinite either.

Member said:
The only exception would be in situations where there
is so much of a particular shade in the horizontal axis, that the
vertical reading goes off the top of the histogram display (which
raraely happens on an image that is reasonably well exposed, unless
there subject is mostly a single shade).

If an image has a lot of one color, that color will be represented as a hump in some length of the histogram. You can see this easily in Photoshop, where you can view the indivisual colors as a separate histogram.

Member said:
On my camera (Dimage 7), the area of the histogram varies
considerably from shot to shot. Can anyone explain why?

If you have a 'high key' image, the tonal values will be mostly in the highlight range, and that will be reflected as such, likewise a 'dramatic' shot will show more values in the lower range.

If you have two identical cameras with indentical exposures the histogram should be the same, but never between two cameras or exposures.

The only practical uses of the histogram I've found are to check the high and low range of the exposure, and to check the integrity of the overall values.

I understand why you need it to confirm your digital exposures if your camera is not tethered and the function is available on the back of the camera, but understanding of the histogram is of little practical use otherwise. I don't even recommend that it be used in Photoshop with levels.

Are we on the same wave lengh?

John G.

Thomas Sapiano · Dec 16, 2001

The problem is that the vertical axis has a dynamic scale that is not in any particular units. It automatically rescales itself to keep the peeks at a reasonable level in the plot, so nothing gets cut off, but subtle details aren't droped out of significance. It isn't meant to be a numerical plot, simply to demonstrate the statistical distribution of the luminoscity in the scene to get a rough idea of the exposure.

If there are large spikes, the plot may be very close to the bottom - making a small volume under the plot. On the other hand, with a relatively even distribution, the plot will have a large volume under the curve, as there is no spikes bumping up the upper level of the range.

Kent said:
As I understand the theory of the histogram, the graph is a
representation of the distribution of grey scale from pure black to
pure white. Colors are converted to a black and white equivalent.

The vertical height of any point is a measure of the number of
pixels (or some other small unit of measure) at that particular
point on the horizontal axis (from black at the left to white at
the right).

Since evey camera at any specific setting has a fixed number of
pixels or units for analysis in generating a histogram, I would
expect the area under the curve of a histgram to be constant for
any camera. The only exception would be in situations where there
is so much of a particular shade in the horizontal axis, that the
vertical reading goes off the top of the histogram display (which
raraely happens on an image that is reasonably well exposed, unless
there subject is mostly a single shade).

On my camera (Dimage 7), the area of the histogram varies
considerably from shot to shot. Can anyone explain why?

Kent
San Francisco
--
Kent

KenEis · Dec 16, 2001

Yes the area under the curve is a constant--for a given camera. Two things change the constant, number of pixels and the depth of the pixels 8-14 bits deep.

Whether you take an all black picture (everything is in bin 0) or all white (everything is in the highest bin) or take a normal image you get the same area under the curve.

Andrew Grant said:
Kent said:

As I understand the theory of the histogram, the graph is a
representation of the distribution of grey scale from pure black to
pure white. Colors are converted to a black and white equivalent.

The vertical height of any point is a measure of the number of
pixels (or some other small unit of measure) at that particular
point on the horizontal axis (from black at the left to white at
the right).

Since evey camera at any specific setting has a fixed number of
pixels or units for analysis in generating a histogram, I would
expect the area under the curve of a histgram to be constant for
any camera. The only exception would be in situations where there
is so much of a particular shade in the horizontal axis, that the
vertical reading goes off the top of the histogram display (which
raraely happens on an image that is reasonably well exposed, unless
there subject is mostly a single shade).

On my camera (Dimage 7), the area of the histogram varies
considerably from shot to shot. Can anyone explain why?

Kent
San Francisco
--
Kent

Click to expand...

DavidMillier · Dec 17, 2001

JohnG said:
I think that you're mixing up terms, and the historgram can be
understood easily for practical work, much less so for speculation.
I know how to use it in Photoshop, and I assume you know how to use
it practically for your applications, so perhaps you want to
rephrase?

Kent's description seemed spot on to me!

john g said:
Kent said:

As I understand the theory of the histogram, the graph is a
representation of the distribution of grey scale from pure black to
pure white. Colors are converted to a black and white equivalent.

Click to expand...

Well, not really a representation of grey scale, it shows all the
colors in a term you might call: brightness, luminosity, value,
density, or % of total ink.

...which is a greyscale...

john g said:
john g said:

The vertical height of any point is a measure of the number of
pixels (or some other small unit of measure) at that particular
point on the horizontal axis (from black at the left to white at
the right).

Click to expand...

Not pixels at all, some other small unit.

it's a measure of the relative distribution of the number of pixels of each greyscale value

john g said:
john g said:

Since evey camera at any specific setting has a fixed number of
pixels or units for analysis in generating a histogram, I would
expect the area under the curve of a histgram to be constant for
any camera.

Click to expand...

Not fixed, but not infinite either.

I would expect the area to be constant for any one model of camera. Presumably the reason they aren't is because the histogram is simplified a bit to fit it on the tiny LCD

john g said:
john g said:

The only exception would be in situations where there
is so much of a particular shade in the horizontal axis, that the
vertical reading goes off the top of the histogram display (which
raraely happens on an image that is reasonably well exposed, unless
there subject is mostly a single shade).

Click to expand...

If an image has a lot of one color, that color will be represented
as a hump in some length of the histogram.

Huh? The in camera histogram displays information about luminance only, there is no colour information present i.e. the histogram for a green card will look identical to that of a grey card if the luminosity of the two is the same.

You can see this easily

john g said:
in Photoshop, where you can view the indivisual colors as a
separate histogram.

john g said:

On my camera (Dimage 7), the area of the histogram varies
considerably from shot to shot. Can anyone explain why?

Click to expand...

If you have a 'high key' image, the tonal values will be mostly in
the highlight range, and that will be reflected as such, likewise a
'dramatic' shot will show more values in the lower range.

If you have two identical cameras with indentical exposures the
histogram should be the same, but never between two cameras or
exposures.

The only practical uses of the histogram I've found are to check
the high and low range of the exposure, and to check the integrity
of the overall values.

I understand why you need it to confirm your digital exposures if
your camera is not tethered and the function is available on the
back of the camera, but understanding of the histogram is of little
practical use otherwise. I don't even recommend that it be used in
Photoshop with levels.

Are we on the same wave lengh?

John G.

Leon Wittwer · Dec 17, 2001

In Photoshop, the histogram just consists of the accumulation of the hits at a RGB coordinate value over the three colors. I don't believe that there is any conversion to grey scale. I suspect that digital cameras likely work the same way if, for no other reason, it reduces the amount of calculation.

john g · Dec 17, 2001

Thinking that Luminosity is the same as Grey Scale is wrong. View the histogram of an image with color information, and then convert it to Grey Scale and view it's histogram. If they are the same it is a coincidence. Let me know what happens.

The real question is why this information is important? I entertained the theoritical question, but the premise and terminology is all wrong.

I understand that with digital cameras with LCD screens you can confirm your exposure, but overall, the histogram and levels are pretty simple, as in Photoshop.

Using the term 'pixels' is confusing and misleading because the histogram is only remotely related to pixels, especially the ones from the camera's CCD.

Anyway, I don't know what to say, I can't see this thread going anywhere, I'm outta here.

John G.

DavidMillier said:
Kent's description seemed spot on to me!

it's a measure of the relative distribution of the number of pixels

DavidMillier said:
of each greyscale value

I would expect the area to be constant for any one model of camera.

DavidMillier said:
Presumably the reason they aren't is because the histogram is
simplified a bit to fit it on the tiny LCD

Huh? The in camera histogram displays information about luminance

DavidMillier said:
only, there is no colour information present i.e. the histogram for
a green card will look identical to that of a grey card if the
luminosity of the two is the same.

Steve Newbold · Dec 17, 2001

There seems to be some misunderstanding about exactly what the histogram is actually showing us.

The histogram basically plots a chart using all the pixels within the image (yes, they are pixels!). It checks the overall luminosity (or Intensity) of the 3 colour channels, this is not actually the grayscale equvalent.

It then plots the number of pixels at each level of intensity (0-255) against an axis. So essentially what we are seeing is a graphical representation of how much detail we have from the deepsest shadow (black) to the brightest highlight (white) with all the midtones in between.

The 'y' axis is always scaled so that the highest peak reaches the top of the scale, regardless of it's actual value be it 5 pixels or 2500 pixels.

Now the bit that if think the first question is about. The histogram graph has the same 'x' scale regardless of the image type (8bit, 16bit, number of pixels etc.) Therefore, each image has a unique histogram regardless of it coming from the same source (because one axis is the same, the only way a histogram can be the same is if the image had exactly the same number of pixels with exactly the same luminosity values, which is astronomically implausible!!!)

The histogram, is therefore, the most useful tool a digital artist/photographer has. It essentially tells us the quality of our image. It tells when we have pushed things to far (expanding the histogram, causing gaps and therefore posterisation) how much detail we have in shadows and highlights. It is essential to obtaing great digital images.

It should also be said that the histogram is useless unless you deselect 'use cache for histogram' in the preferences in Photoshop, as it will only use a sample of the image for the histogram rather than looking at every pixel in the image.

I hope this has helped,

Steve Newbold.

Dick- · Dec 17, 2001

Wrong, John's explanation is accurate..otherwise the histogram would be of no value. dm

DMillier said:
JohnG said:

I think that you're mixing up terms, and the historgram can be
understood easily for practical work, much less so for speculation.
I know how to use it in Photoshop, and I assume you know how to use
it practically for your applications, so perhaps you want to
rephrase?

Click to expand...

Kent's description seemed spot on to me!

DavidMillier said:

Kent said:

As I understand the theory of the histogram, the graph is a
representation of the distribution of grey scale from pure black to
pure white. Colors are converted to a black and white equivalent.

Click to expand...

Well, not really a representation of grey scale, it shows all the
colors in a term you might call: brightness, luminosity, value,
density, or % of total ink.

Click to expand...

...which is a greyscale...

DavidMillier said:

DavidMillier said:

The vertical height of any point is a measure of the number of
pixels (or some other small unit of measure) at that particular
point on the horizontal axis (from black at the left to white at
the right).

Click to expand...

Not pixels at all, some other small unit.

Click to expand...

it's a measure of the relative distribution of the number of pixels
of each greyscale value

DavidMillier said:

DavidMillier said:

Since evey camera at any specific setting has a fixed number of
pixels or units for analysis in generating a histogram, I would
expect the area under the curve of a histgram to be constant for
any camera.

Click to expand...

Not fixed, but not infinite either.

Click to expand...

I would expect the area to be constant for any one model of camera.
Presumably the reason they aren't is because the histogram is
simplified a bit to fit it on the tiny LCD

DavidMillier said:

DavidMillier said:

The only exception would be in situations where there
is so much of a particular shade in the horizontal axis, that the
vertical reading goes off the top of the histogram display (which
raraely happens on an image that is reasonably well exposed, unless
there subject is mostly a single shade).

Click to expand...

If an image has a lot of one color, that color will be represented
as a hump in some length of the histogram.

Click to expand...

Huh? The in camera histogram displays information about luminance
only, there is no colour information present i.e. the histogram for
a green card will look identical to that of a grey card if the
luminosity of the two is the same.

You can see this easily

DavidMillier said:

in Photoshop, where you can view the indivisual colors as a
separate histogram.

DavidMillier said:

On my camera (Dimage 7), the area of the histogram varies
considerably from shot to shot. Can anyone explain why?

Click to expand...

If you have a 'high key' image, the tonal values will be mostly in
the highlight range, and that will be reflected as such, likewise a
'dramatic' shot will show more values in the lower range.

If you have two identical cameras with indentical exposures the
histogram should be the same, but never between two cameras or
exposures.

The only practical uses of the histogram I've found are to check
the high and low range of the exposure, and to check the integrity
of the overall values.

I understand why you need it to confirm your digital exposures if
your camera is not tethered and the function is available on the
back of the camera, but understanding of the histogram is of little
practical use otherwise. I don't even recommend that it be used in
Photoshop with levels.

Are we on the same wave lengh?

John G.

Click to expand...

--****

DavidMillier · Dec 17, 2001

JohnG

Thanks for the comments - they sent me back to the manual to check!

The full manual for my E10 has the following to say about the histogram:

"During playback you can display a histogram, a bar graph which illustrates the brightness of the displayed image so you can determine how to adjust exposure compensation. "

Note the word "brightness" - no comment about colour here...

"The horizontal axis of the histogram represents the brightness of the image, on a scale of 0~255 (reading from left to right), and the vertical axis measures the number of pixels for each bar. "

...Again Oly seem fairly confident it is measuring brightness and counting pixels...

" If the graph shows brightness heavily weighted on the right side of the scale (portions of the image could wash out), set exposure compensation to a minus value for a darker image. If darkness is heavily weighted on

the left side of the scale (portions of the image could turn black), set exposure compensation to positive value for a brighter image."

...Oly also seem to think it serves a useful function for fine tuning exposure.

I would add a second function too - contrast control. The E10 offers 3 settings for contrast: high, normal, low. If you are very fussy you can change the contrast setting to suit the subject brightness range by snapping a shot then checking the histogram. For a low brightness range scene the histogram will not use the full range 0-255 so you have the option of expanding that by setting contrast to high. And vice versa for high contrast scenes.

JohnG said:
Thinking that Luminosity is the same as Grey Scale is wrong. View
the histogram of an image with color information, and then convert
it to Grey Scale and view it's histogram. If they are the same it
is a coincidence. Let me know what happens.

Difficult this one, as my comments are restricted to the in camera histogram and the E10 doesn't do in camera black and white...but I'll try it out on the histograms in a couple of editing packages (I'll have to take your word about PS as I don't have it).

john g said:
The real question is why this information is important? I
entertained the theoritical question, but the premise and
terminology is all wrong.

I don't understand why you think the terminology is wrong - which words?

I also don't understand which premise you believe is wrong? The idea that the histogram is useful? If so, I think the premise is right, I obviously can't apply the above to cameras I don't use but for the E10 the histogram appears to offer considerable value for field use.

john g said:
I understand that with digital cameras with LCD screens you can
confirm your exposure, but overall, the histogram and levels are
pretty simple, as in Photoshop.

Using the term 'pixels' is confusing and misleading because the
histogram is only remotely related to pixels, especially the ones
from the camera's CCD.

The Oly manual doesn't agree...

john g said:
Anyway, I don't know what to say, I can't see this thread going
anywhere, I'm outta here.

I think perhaps you are prematurely dismissing a useful photgraphic control here.

I'll get back on that brightness/colour thing.

Regards

Dave Millier

john g said:
John G.

john g said:

Kent's description seemed spot on to me!

Click to expand...

it's a measure of the relative distribution of the number of pixels

john g said:

of each greyscale value

Click to expand...

I would expect the area to be constant for any one model of camera.

john g said:

Presumably the reason they aren't is because the histogram is
simplified a bit to fit it on the tiny LCD

Click to expand...

Huh? The in camera histogram displays information about luminance

john g said:

only, there is no colour information present i.e. the histogram for
a green card will look identical to that of a grey card if the
luminosity of the two is the same.

Click to expand...

DavidMillier · Dec 17, 2001

Would you mind explaining in more detail?

Thanks

Dave Millier

****- said:
DMillier said:

JohnG said:

I think that you're mixing up terms, and the historgram can be
understood easily for practical work, much less so for speculation.
I know how to use it in Photoshop, and I assume you know how to use
it practically for your applications, so perhaps you want to
rephrase?

Click to expand...

Kent's description seemed spot on to me!

Member said:

Kent said:

As I understand the theory of the histogram, the graph is a
representation of the distribution of grey scale from pure black to
pure white. Colors are converted to a black and white equivalent.

Click to expand...

Well, not really a representation of grey scale, it shows all the
colors in a term you might call: brightness, luminosity, value,
density, or % of total ink.

Click to expand...

...which is a greyscale...

Member said:

Member said:

The vertical height of any point is a measure of the number of
pixels (or some other small unit of measure) at that particular
point on the horizontal axis (from black at the left to white at
the right).

Click to expand...

Not pixels at all, some other small unit.

Click to expand...

it's a measure of the relative distribution of the number of pixels
of each greyscale value

Member said:

Member said:

Since evey camera at any specific setting has a fixed number of
pixels or units for analysis in generating a histogram, I would
expect the area under the curve of a histgram to be constant for
any camera.

Click to expand...

Not fixed, but not infinite either.

Click to expand...

I would expect the area to be constant for any one model of camera.
Presumably the reason they aren't is because the histogram is
simplified a bit to fit it on the tiny LCD

Member said:

Member said:

The only exception would be in situations where there
is so much of a particular shade in the horizontal axis, that the
vertical reading goes off the top of the histogram display (which
raraely happens on an image that is reasonably well exposed, unless
there subject is mostly a single shade).

Click to expand...

If an image has a lot of one color, that color will be represented
as a hump in some length of the histogram.

Click to expand...

Huh? The in camera histogram displays information about luminance
only, there is no colour information present i.e. the histogram for
a green card will look identical to that of a grey card if the
luminosity of the two is the same.

You can see this easily

Member said:

in Photoshop, where you can view the indivisual colors as a
separate histogram.

Member said:

On my camera (Dimage 7), the area of the histogram varies
considerably from shot to shot. Can anyone explain why?

Click to expand...

If you have a 'high key' image, the tonal values will be mostly in
the highlight range, and that will be reflected as such, likewise a
'dramatic' shot will show more values in the lower range.

If you have two identical cameras with indentical exposures the
histogram should be the same, but never between two cameras or
exposures.

The only practical uses of the histogram I've found are to check
the high and low range of the exposure, and to check the integrity
of the overall values.

I understand why you need it to confirm your digital exposures if
your camera is not tethered and the function is available on the
back of the camera, but understanding of the histogram is of little
practical use otherwise. I don't even recommend that it be used in
Photoshop with levels.

Are we on the same wave lengh?

John G.

Click to expand...

Click to expand...

--
****

DavidMillier · Dec 17, 2001

Steve

Everything you say seems exactly in accord with my understanding except the comment about the brightness of the 3 channels being different from the greyscale. Can you explain this in more detail?

Thanks

Dave Millier

Steve Newbold said:
There seems to be some misunderstanding about exactly what the
histogram is actually showing us.

The histogram basically plots a chart using all the pixels within
the image (yes, they are pixels!). It checks the overall
luminosity (or Intensity) of the 3 colour channels, this is not
actually the grayscale equvalent.

It then plots the number of pixels at each level of intensity
(0-255) against an axis. So essentially what we are seeing is a
graphical representation of how much detail we have from the
deepsest shadow (black) to the brightest highlight (white) with all
the midtones in between.

The 'y' axis is always scaled so that the highest peak reaches the
top of the scale, regardless of it's actual value be it 5 pixels or
2500 pixels.

Now the bit that if think the first question is about. The
histogram graph has the same 'x' scale regardless of the image type
(8bit, 16bit, number of pixels etc.) Therefore, each image has a
unique histogram regardless of it coming from the same source
(because one axis is the same, the only way a histogram can be the
same is if the image had exactly the same number of pixels with
exactly the same luminosity values, which is astronomically
implausible!!!)

The histogram, is therefore, the most useful tool a digital
artist/photographer has. It essentially tells us the quality of
our image. It tells when we have pushed things to far (expanding
the histogram, causing gaps and therefore posterisation) how much
detail we have in shadows and highlights. It is essential to
obtaing great digital images.

It should also be said that the histogram is useless unless you
deselect 'use cache for histogram' in the preferences in Photoshop,
as it will only use a sample of the image for the histogram rather
than looking at every pixel in the image.

I hope this has helped,

Steve Newbold.

DavidMillier · Dec 17, 2001

Steve

In response to your statement (and the similar comment earlier in the thread by JohnG):

Steve Newbold said:
It checks the overall
luminosity (or Intensity) of the 3 colour channels, this is not
actually the grayscale equivalent.

... I find this very confusing (in the sense that it contradicts my 'understanding' ).

I have loaded a colour shot into Picture Window 3 (my preferred editor) and examined its brightness curve in the brightness curve transformation window;

I also converted it into a 16bit B&W and compared the curves;

and they are identical.

This transformation is not a histogram as such (it shows the curve of the highest points of the histogram) but I believe it is equivalent.

Unfortunately, I don't have PS and the histogram in PaintShopro shows the R, G, B or greyscale as separate curves so I can't see another way of viewing a combined RGB.

On the other hand I thought that was exactly what the greyscale was - the average of the RGB curves!!

If not can you explain the difference??

Regards

Dave Millier

Steve Newbold · Dec 17, 2001

OK, this is gonna get pretty technical and is pretty much irrelevant but here goes...

There are a number of different models that describe colour. RGB, CMYK, LAB, HSB etc.

Normally we only work in RGB (eg. light) or CMYK (eg. ink) modes. But behind the scenes, Photoshop and other programs work in different colour modes for different functions.

We are concerened with the Histogram function. Your question is how the 'luminosity' and 'grayscale' differ when at first they seem to be identical in Histogram.

First of all we need to define exactly what a grayscale image is, and what luminosity means. When an RGB image is converted into a grayscale all that is happening is that the program gets rid of 2 of the channels in the HSB model, namely the 'H', hue, or colour, and 'S', saturation. These two channels describe what colour the pixel is, and how saturated it is (completely unsaturated to fully saturated.)

All that is left is the 'B' channel, brightness. This describes relative brigtness of the pixel, from white to black. If we get rid of all the colour information all we are left with is this image made up of relative brightness, a grayscale image. It is the same as averaging all the values of 3 channels in RGB. You get the same result.

On the other hand Luminosity comes from yet another colour model, LAB. The way this model describes colour is by having 3 channels.

Luminostity, that describes the lightness of a pixel, 'a' a chromatic component from green to red, and, 'b' a chromatic value from blue to yellow. When combined these channels give exactly the same colours as their RGB or HSB counterparts, its just a different way of getting there. Because of the different values of the other two channels compared to the HSB model, the luminosity channel HAS to be different to a Brightness channel or grayscale average. Altgough they do look quite similar to the eye on screen and in a Histogram.

Again this is long winded, overly-complicated and irrelavnt but it is the reason.

I should be able to prove it using these images.

This is a test image used to show the difference. The image was duplicated. I ran a Histogram on the first image as it is. (below)

This Histogram shows the Luminosity values of each pixel in the RGB image.

Next I converted the other version to grayscale and performed another histogram. A histogram of a grayscale image show only the grayscale values of each pixel the image. Note the subtle differences between the histograms.

Hope that clears things up,

Steve

DavidMillier · Dec 17, 2001

Steve

Thanks for the explanation - but I doubt if you seriously thought it would clear things up

If you don't mind, I've a couple of comments/questions...

I appreciate that there are different ways of encoding information about colour but if I remember my physics/biology/pyschology there is only one way of detecting/perceiving colour! The eye has 3 different colour receptors sensitive to a range of different wavelengths (roughly rgb).

When reflected or transmitted coloured light hits the retina, the relative proportions of the wavelengths are somehow calculated and you perceive colour. So colour perception is essentially a RGB additive system.

Full spectrum light meters for photographic use are designed (as carefully as possible at least) to respond to brightness and colour somewhat like film, but the measured results are averaged so that they simply report the overall amount of light reflected from the subject (traditionally in EV values).

Now I have always considered this to be what is meant by a greyscale i.e. a simple reduction of the subject to brightness values.

And this is exactly what I see the histogram on my E10 as doing i.e when the camera records a file as a set of RGB values for each pixel, the histogram function adds the values of the RGB components of each pixel and divides by 3 to yield an average grey brightness for each pixel. It then plots the relative distibution of these pixel brightnesses on the hostogram display.

Consequently, there is no colour information whatsoever in the in-camera histogram display, just relative greyscale brightnesses. In other words, the in camera histogram pretends the camera is a black and white camera only and the histogram is a bit like a refined Ansel Adams zone map of subject brightnesses.

Have I got this completely wrong?

Steve Newbold · Dec 18, 2001

Luminosity is a 'greyscale' representation of brightness, it's just not the same as averaging out the tones of the RGB channels. Its like one is the accurate way of doing it, the other is the cruder way of doing it.

When you actually look at the difference, it may only be that the mean difference is 0.1 different. This means that the deviation between the two modes may be less than a 10th brighter/darker throughout the range of the image.

I don't know why Photoshop (and other Pro software) do it in this way but they do. It may well be the case that your camera does use the grayscale version to plot its histogram. Eitherway, the result would be so similar there is no risk of you altering exposure because of it.

How the eye percieves brightness depending on the light levels present, low light it uses 'rods' to measure light intensity (but not colour), and in normal light it uses a combination of the three light sensetive cones (as you sate) to percieve colour. A combination of these two methods is used most of the time whicj is why we can pick out such detail in shadows and highlights as well as subtle shades of colour. Any machine that measures light has to be crude in comparison.

As I said before, it doesn't really make much difference. I'm sure that there is no standard between cameras and software to produce histograms as it amkes no difference. I DO feel that it is an amazing tool for looking at images in a different way though.

Steve Newbold · Dec 18, 2001

Just thought of something that may make the difference more obvious.

Going back to the eye. As I said the eye has two ways of percieving light.

The first, the rods, are only sensitive to the intensity of light. They only see if it light or dark or anywhere in between. This is the equvalent of the luminosity of an image.

Know if we look at the cones, each cone is either sensitive to roughly red, green or blue. So somehow in our brain we combine these three images to percieve colour. A typical grayscale version of an image is the average of these three channels. But the brightness of these three channels is different, ie. the red channel sees skin as light but the sky as dark, but the blue channel sees the sky as light etc. So when we average these channels we are getting slight variations that wouldn't be present in the luminosity channel.

Steve

DavidMillier · Dec 18, 2001

Thanks for this Steve

I'm not entirely sure I completely understand the reasons why greyscale is not quite the same as average RGB but from what you say the differences are negligible so I'm happy to ignore the differences.

The significant thing we are left with is that the in camera histogram is a potentially very useful tool! Some people seem to think this is not true and they are missing out on a photographic control.

I only recently got into digital photography after a number of years of doing my own b&W large format stuff (hobby, not pro).

One thing that has intrigued me on the E10 is the contrast control - buried in the menus sadly. The general advice on the Oly forums is to stick it on Normal (or possibly Low) and leave it there. Zone system background however makes me interested in this option. From running some tests and looking at the differences on the in-camera histogram, I reckon the contrast control should be used to match the camera's dynamic range to the subject brightness range and should be treated as a photographic control rather than set and forget. The histogram is useful for making decisions on this in the field.

Regards

Dave Millier

ps

An aside I know, but I once spent a couple of sessions in the pub with a friend in which we gave some thought to the eye's mechanism for perceiving colour.

We came up with the idea (simply by extrapolating text book descriptions and drinking lots of beer), that the eye/brain uses a redundant system whereby there are sometimes more that one combination of RGB that yields the perception of a particular colour. Not true for all colours, but yellow could be one example. A pure monochromatic yellow will produce the perception of the colour yellow but so will the combination of blue and green light. There must be other colours too that this will work for.

Steve Newbold said:
Just thought of something that may make the difference more obvious.

Going back to the eye. As I said the eye has two ways of
percieving light.
The first, the rods, are only sensitive to the intensity of light.
They only see if it light or dark or anywhere in between. This is
the equvalent of the luminosity of an image.

Know if we look at the cones, each cone is either sensitive to
roughly red, green or blue. So somehow in our brain we combine
these three images to percieve colour. A typical grayscale version
of an image is the average of these three channels. But the
brightness of these three channels is different, ie. the red
channel sees skin as light but the sky as dark, but the blue
channel sees the sky as light etc. So when we average these
channels we are getting slight variations that wouldn't be present
in the luminosity channel.

Steve

Frank Benvenuto · Dec 18, 2001

This was a very interesting thread. I thank Steve and Dave for the excellent posts.. A bit over my head but hey... it was good reading... Now for the $1,000,000 question. HOW do you utilize histograms to help make better exposures? I am an old film buff that has recently jumped onto the Digital bandwagon with a D1x. I want to learn about Histograms as best I can.

DMillier said:
I'm not entirely sure I completely understand the reasons why
greyscale is not quite the same as average RGB but from what you
say the differences are negligible so I'm happy to ignore the
differences.

The significant thing we are left with is that the in camera
histogram is a potentially very useful tool! Some people seem to
think this is not true and they are missing out on a photographic
control.

I only recently got into digital photography after a number of
years of doing my own b&W large format stuff (hobby, not pro).

One thing that has intrigued me on the E10 is the contrast control

buried in the menus sadly. The general advice on the Oly forums

is to stick it on Normal (or possibly Low) and leave it there. Zone
system background however makes me interested in this option. From
running some tests and looking at the differences on the in-camera
histogram, I reckon the contrast control should be used to match
the camera's dynamic range to the subject brightness range and
should be treated as a photographic control rather than set and
forget. The histogram is useful for making decisions on this in
the field.

Regards

Dave Millier

ps

An aside I know, but I once spent a couple of sessions in the pub
with a friend in which we gave some thought to the eye's mechanism
for perceiving colour.

We came up with the idea (simply by extrapolating text book
descriptions and drinking lots of beer), that the eye/brain uses a
redundant system whereby there are sometimes more that one
combination of RGB that yields the perception of a particular
colour. Not true for all colours, but yellow could be one example.
A pure monochromatic yellow will produce the perception of the
colour yellow but so will the combination of blue and green light.
There must be other colours too that this will work for.

Steve Newbold said:

Just thought of something that may make the difference more obvious.

Going back to the eye. As I said the eye has two ways of
percieving light.
The first, the rods, are only sensitive to the intensity of light.
They only see if it light or dark or anywhere in between. This is
the equvalent of the luminosity of an image.

Know if we look at the cones, each cone is either sensitive to
roughly red, green or blue. So somehow in our brain we combine
these three images to percieve colour. A typical grayscale version
of an image is the average of these three channels. But the
brightness of these three channels is different, ie. the red
channel sees skin as light but the sky as dark, but the blue
channel sees the sky as light etc. So when we average these
channels we are getting slight variations that wouldn't be present
in the luminosity channel.

Steve

Click to expand...

--Frank B

Histograms - Theory Question

Kent

Guest

Senior Member

Leading Member

Senior Member

Senior Member

Forum Pro

Forum Pro

Leading Member

Member

Dick-

Guest

Forum Pro

Forum Pro

Forum Pro

Forum Pro

Member

Forum Pro

Member

Member

Forum Pro

Leading Member

Similar threads

Keyboard shortcuts