Why the difference in Levels on Photo Restorations?

[dennford]

Sorry Kieth, also forgot to add that if you have done a lot of work such as colour editing it can also affect the continuety of one or more colours, thereby having the same effect.

                                              Denn

[dennford]

Just out of interest you may want to look at this site for an explanation of levels - dont be put off by the title it has an explanation of levels that is very easy to understand. it is from a site called photoshop for astrophotographers - sounds grand and very cerebral but in fact is a simple site (should be more like it)

http://www.astropix.com/PFA/SAMPLE3/SAMPLE3.HTM

                                    Denn

[PrueM]

Here is another explanation of levels that I found useful - describes it much better than I ever could!

http://www.earthboundlight.com/phototips/photoshop-levels-1-2-3.html

[Keith Bateman]

Thanks Denn & Prue - those pages explain a lot.

I knew  what happened when I moved the sliders etc - but never knew why it was happening.

The simple coins diagram explains everything.

Only trouble now is I will have no excuse for not making all photos better - but of course I can always blame the scanner!! 

Cheers

Keith

[dennford]

Actually Kieth, if the coins diagram had have shown the gaps it may well have been describing my houshold budget!

                                                        Denn

[tomkin]

Keith,
          I suspect that it is all to do with adjusting High Bit Depth Images.

     Tomkin

[arnietheother]

Hello,
I have a theory.
I've seen this effect before on digitally edited images.  I think it's caused by expanding the range of a channel with levels / curves or similar tools. 
I'm not sure what the RGB channel is, perhaps some sum of the R, G & B channels? My understanding is that it's some representation of the distribution of values from black to white in the image. 
My theory is that the channel starts out as a smoothly varying distribution of levels, but as the range of a channel is expanded, pixel levels are re-distributed up or down, but with a weighting, probably due to integer truncation or some other software imperfection.  So over a very narrow range of the graph, most or even all values will be rounded down and hardly any or none rounded up, resulting in the skewed distribution we see in the RGB channel graph. 
Plausible?

Arnie.

[tomkin]

Very near,

Tomkin

[dennford]

All this apart, here is a practical tip for levels adjustment that can make it more effective.
Your rgb histogram gives an indication of what levels are present in all three colours, but they are rarely all the same, so in your levels box is the ability to choose either r,g or b. Now if you adjust the levels of each one seperately, the result will be different to that of adjusting them as a whole - usually a better result.