Histograms - a treatment for cold sores?

XEyedBear

Suspended / Banned
Messages
12
Name
Tony
Edit My Images
Yes
I'm currently reading 'Camera RAW for Digital Photographers Only' by Rob Sheppard - a book which I can with all sincerity thoroughly recommend that you ignore (that's another forum, I guess). He talks (God, does he talk..) about Histograms. If you have a copy of this book, look at page 26. Three histograms are shown of which the top 2 have intrigued me.

These two histograms show the data from two reasonable well exposed images. There is no clipping and no extreme, or dominating, peaks. The area 'below the curve' in one histogram occupies about 70% of the available graph area, the other occupies only 30%.

Assuming these are 2 images from the same sensor, how can this be?

If I understand them correctly, in a histogram each point on the x-axis represents one of the 'quantisation' levels that each sensor (or it group of 4 sensors: GRGB ?) is capable of measuring. I guess there are 2^8 quantisation levels for each sensor for a total of 24 bits of colour tone (assuming only 1 of the 2 'green' sensors in each pixel is used for this purpose).

The Y-axis is a measure of the number of sensors that have measured each of these 2^8 quantisation levels. So the total of all the y-axis readings should equal the total number of sensors points on the light measuring device.
And therefore for 2 images from the same sensor the sum (or integral) of the y-axis readings should be constant. But for the two figures in this book they are demonstrably different.

What have I misunderstood?
 
....errr........what can I say???? Help me out ......somebody......please......
 
Cartier-Bresson just turned in his grave.

Sorry I have absolutely no idea what you are on about !
 
Damn! page 26 missing from my copy, sorry can't help :shrug:
 
I'm going to reroute the power conduits through the aft flux couplers which should stabilise the quantum matrix particle transfer drives. Hopefully.....
 
And therefore for 2 images from the same sensor the sum (or integral) of the y-axis readings should be constant. But for the two figures in this book they are demonstrably different.

What have I misunderstood?

I would have expected the definite integral to be the same, too. Perhaps the cameras displaying the histograms have rescaled the Y-axis?
 
What have I misunderstood?

Well, if I understood your post then you've mixing up what the histogram represents. The x axis is the brightness level either for grey or rgb starting with the darkest on the left and the brightest on the right.

The Y axis shows how much of the image is using the brightness level at X. So if you had an image with 10 pixels with a luma of 15 then at x=15 y would be 10, it's that simple.

If you take a picture of a dark scene then the historgram will show a lot of height on the left and very little or nothing on the right. An under exposed shot would be "clamped" to the left, where the leftmost point will be all the way to the top indicating that you've got an excess of 0 value data in the image.

A bright scene will be the exact opposite, lots on the right and not much on the left. An over exposed shot will be "clamped" to the right indicating an excess of 255 (in 8bit) value data.

A well exposed, average scene will show a gentle curve from left to right with the heighest point in the middle.

Hopefully now you can understand why the heights shouldn't always add up to the total number of sensors because the graph doesn't represent all of the sensor points, it represents the distribution of values from the image.
 
Well, if I understood your post then ........
..........................................
...........it represents the distribution of values from the image.

:bang: Exactly!!!! That's what we've all been trying to say here!!! :notworthy: :thinking:
 
Back
Top