How does equalize handle color?

[CM Jamieson]

I understand how histogram equalization works, but almost all online material on equalization talks about the grey-scale case.

Extending this to the colour case is not obvious and several different approaches are available.

So, how exactly does equalize handle color?

[George Austin]

CM

I have looked at the Equalize feature and have determined this much:

In the selected area, Equalize finds the highest and lowest color values in any of the three RGB channels. It changes those color values to 255 and 0, respectively, in the channel and at the pixel site where found. The PS manual and other references I've found confuse this issue by saying the conversion is to "white" and "black".

A value of 255 in a single channel with the other channels unspecified is a far cry from white, just as 0 in a single channel with the other channels unspecified is a far cry from black. So "black" and "white" are mis-used in the literature on Equalize.

I have not yet figured out how, after pushing the highest and lowest values to their extremes in separate channels, Equalize then distributes values in all three channels, but at least I think I've got this one point of confusion untangled.

George

[Chris Cox]

Equalize distributes the values to equalize the channel histograms (it's a statistics thing).

Yes, it works per channel.

[George Austin]

"...to equalize the channel histograms..."

Clear as mud, Chris.

Does equalize mean the red and green and blue histograms are somehow "equal"? They sure don't look alike.

Or does it mean that all the spikes in, say, the red histogram are equal in height, and likewise in the green and blue histograms? No, that doesn't fly.

You've got me. I don't know what "equalized" histograms means.

George

[Chris Cox]

"Equalize" is a term from statistics. (spread out evenly)

Again, all the math is PER CHANNEL - not between channels.

[George Austin]

Chris,

....all the math is PER CHANNEL - not between channels..."

I am having difficulty reconciling that groundrule with the observed cross-channel dependence when all pixels are set so that everywhere:

R > 160

160 > G > 80

B < 80

With those values, the lowest red is higher than the highest green anywhere and the lowest green is higher than the lowest blue anywhere, so there is never the slightest doubt which channel is high, intermediate, and low.

If the red and blue channels are left alone and the green channel is altered, but still conforming to the condition that 160 > G > 80, the red and blue values after invoking the equalize feature change as the green channel is varied.

If the intermediate channel is affecting the high and low channels and there is no overlapping of high, intermediate, and low channels anywhere, how can we say the math is per channel and not between channels.

George

[Chris Cox]

There is NO cross channel dependence in the equalize command.
The inter-channel relationships don't matter at all.
It's equalizing each channel completely independently.
How many ways do I have to say this?

[George Austin]

I will not ask Chris twice

I will not ask Chris twice

I will not ask Chris twice

I will not ask Chris twice

I will not ask Chris twice

I will not ask Chris twice

I will not ask Chris twice

...........................

[George Austin]

Dave,

What, with the clipboard---Piece o' cake!

It was different back in the days of the blackboard...

Another form of penance rendered useless.

George

[Bill Lamp]

* writes a basic program* copy con IWONT.bas
5 x=1
10 If x>10e+39 Then goto 50 else goto 20
20 print "I will not ask Chris twice"
30 x=x+1
40 goto 10
50 END

GAK, it has been way too long since I did any of this.

Bill