Understanding histogram

[Lady GooGoo La La]

I understand when looking at the luminance histogram that if the histogram begins at a level above 0 on the far left then probably shadows are clipped and if above 0 on the far right the highlights are clipped, but what about the histogram going off the top of the graph in the middle?

My guess is that if all channels equally affected then OK, but if any RG or B channel goes off the top, then those pixels will have the wrong tone or tint since one or more channels is/are under-represented.

Is this correct?, If so will decreasing the over-saturated channels correct the problem?

[© Tom Cooper not going to Views]

No, that is not a valid assumption.  If pixels go off the top, then the histogram chart itself is not tall enough, and that is all it means.  This happens when an image has low contrast (overall channel), or is strong in one color (such as mostly sky with no clouds, would be very high in a narrow band on the blue channel).

Think of the height of the parts of the histogram as a count of pixels, because that is all it is.

Tom

[Lady GooGoo La La]

So this is not a hardware limitation!

That seems silly to me, it the histogram was off chart, why wouldnt it normalize itself to the highest peak? Unless if it did that then the rest of the histogram might appear almost flat until the normalized spike. Obviously, most histograms don't normalize.

I had assumed saturation of the sensor or the A/D convertor.

So what you are saying is that no information is lost if spikes over the top, so image should look normal!

OK thanks for that, I was thinking I may need an ND filter, ISO, aperture or shutter speed to bring these spikes on chart.