Clip Indicator, aka the anti-histogram

[stevefal]

I find that when looking at a histogram, the information I'm typically trying to get is expressed with the barest sliver of pixels, so I set out to design an indicator that exaggerates the extremes that I care about.

The goal was to show any number of near-clipping pixels within two stops in both directions, the fraction of clipping pixels, and for all of luminance, R,G & B.

If this is discernible, I shouldn't have to explain much, so here goes:

Clip Indicator (200%)

Details regarding thresholds and other logic are TBD. Thoughts welcome.

[Audionut]

Just to be clear, that's showing clipping at the black and white points with 2EV on the insides?

I'd probably make the white (maximum pixel saturation) point more sensitive.  0.5 EV increments.
And a better way to represent how many pixels are being clipped at either side.  Possibly the number of pixels in thousands.  1K clipped, 2K clipped etc.

[a1ex]

Absolute counting is variable, depends on image size. On the current histogram, overexposure warnings show percentage (1% clipped, 2% and so on; what's smaller than 1% is variable radius, without text).

[stevefal]

Just to be clear, that's showing clipping at the black and white points with 2EV on the insides?

Yep, exactly.

On the current histogram, overexposure warnings show percentage

I was thinking the same thing, with potentially a logarithmic scale. The clip areas are 38 pixels wide, so there's a lot  to play with. However I personally don't think numeric readout should be necessary. The goal was to have something that, with a little bit of experience, you could dart to it and know exactly what's going on.

In the clipping area, I envisioned that a little sliver was probably ok (<0.5%), and that solid was very bad (at least >10%).

[Audionut]

On the current histogram

Ah yes, It's a good representation too.

In that case, I'd shorten the clipping areas (with the same indicators as current to make it more user friendly), which will enable more room for more sensitive representation of the areas close to clipping.

edit:  The current histogram already shows a very good representation of clipping (on the highlight side anyway), so imo, a new histogram should concentrate on different specifics.

[stevefal]

...which will enable more room for more sensitive representation of the areas close to clipping.

I was thinking that the +/- 1-2 EV indicators would light up completely if there were virtually any pixels in that range. The idea was to get away from splitting hairs and make it very easy to understand where the tail of the curve is.

Along those lines don't you figure 1 stop resolution is sufficient?

[Audionut]

Along those lines don't you figure 1 stop resolution is sufficient?

More resolution in the indicators would be handy.  So for instance, instead of having the 1 indicator light up completely with any pixels, have it light up progressively as more and more pixels reach that EV.  I'd probably go as far as having the numeric go bold when that EV has been filled with pixels of any channel.

It's easy for me of course, I'm not the one doing the coding.  I'm a code monkey at best 

edit:  On the highlight side, as pixels reach -1EV (from saturation), the indicator begins to fill up.  As it gets closer to 0EV, more and more of the indicator fills up.

Does that make sense?  I'm typing it and not even sure I can makes sense of it 

[stevefal]

It makes sense but it sorta defeats my point by turning it into a kind of histogram again. For example, it's possible that there's a spike at the top and a dip at, say, -2EV. In your model the -2EV indicator would show low while clipping may still be occurring above that. I was shooting for something simpler that does not try to deal so much with distribution of pixels within buckets, but rather a quick way to see whether exposure thresholds are being touched at all (which is what I usually do with a histogram).