Optimal histogram width

[eonasick]

I use Adobe Camera Raw to process RAW files. I understand ETTR and issues about the histogram exceeding its bounds. My question is: should the histogram always be tweaked in ACR so that it occupies the full width available (i.e. 0,0,0 black to 255,255,255 white)?

For example, consider a low-contrast shot like trees in a fog. The histogram is more or less a spike in the middle gray area. So do I use the sliders in ACR into spread the histogram to full width, and then adjust with brightness and contrast and curves? Or is it okay to leave out a whole slew of blacks? This is a technical, not artisitic question. Thanks

[Tim Gray]

There is no technical reason to have a 0 0 0 black point and 255 255 255 white point.  Optimal shape of the histogram depends on the aesthetic content of the image.

[Diapositivo]

I use Adobe Camera Raw to process RAW files. I understand ETTR and issues about the histogram exceeding its bounds. My question is: should the histogram always be tweaked in ACR so that it occupies the full width available (i.e. 0,0,0 black to 255,255,255 white)?

For example, consider a low-contrast shot like trees in a fog. The histogram is more or less a spike in the middle gray area. So do I use the sliders in ACR into spread the histogram to full width, and then adjust with brightness and contrast and curves? Or is it okay to leave out a whole slew of blacks? This is a technical, not artisitic question. Thanks
[a href=\"index.php?act=findpost&pid=145733\"][{POST_SNAPBACK}][/a]

If the esthetical result is what you want, it is technically the right thing to leave the histogram as it is, showing a spike in the middle gray area.

If you instead "close" the white and black point toward the "spike" so as to spread the histogram, you will increase contrast up to a point where it is too much for the effect you wanted.

Now that you have too much contrast, if you use "Curves" you only redistribute contrast within the same black and white points, so you can diminish contrast in the middle tones but you will have it increased in the lowlights and in the highlights (I am assuming you don't move the extremities of the curves).

If you use "Contrast" or you use Curves but move (towards the center) the extremities of the Curve, you are just undoing what you have previously done while closing (in Levels) the black and white points around the spike.

If your image is low contrast (such as you have in fog, rain, smoke etc) it is normal and technically OK to have a spike in the middle of the Levels histogram, just like it is technically OK to have in on the left if it is a "low-key" image and on the right if it is a "high-key" image.

There is no technical requirement to have your levels spanning the entire range of values.

The reason why it is generally advised to have it is because, for normal images, having the black and white points near the extremities of the Levels histograms (but not clipping) is a way to have a nice contrasty and saturated picture or as they say "to give it some pop". If you go on moving the black and white point arrows toward the center, past the borders of your histogram, you will be increasing your contrast even more, but you will have clipped some lowlights and some highlights. So again it depends on the image, with certain images it is perfectly OK, and even advisable, to clip some black or white portions, but you normally want to have let's say around the 98% of your entire dynamic range unclipped (meaning: if you have 1 or 2 % of shadows in the dark parts of a bush, you let it go to black and that is even better, gives more contrast to the bush, by the same token if you are scanning a nocturn picture the scanner might produce an image which comprises all the highlights, the streetlamps, unclipped. That will probably be not enough contrasted, you will let the streetlamps to clip and increase contrast).

Cheers
Fabrizio

[Jonathan Wienke]

Unless you have a foggy scene, you generally want  your black point near 0,0,0 and white point near 255,255,255. Since prints and displays are already lower-contrast than real life, you generally want to have the entire contrast range to work with. But if you have a foggy image, the fog ambience can be ruined by stretching contrast too much.

[eonasick]

Unless you have a foggy scene, you generally want  your black point near 0,0,0 and white point near 255,255,255. Since prints and displays are already lower-contrast than real life, you generally want to have the entire contrast range to work with. But if you have a foggy image, the fog ambience can be ruined by stretching contrast too much.
[a href=\"index.php?act=findpost&pid=146147\"][{POST_SNAPBACK}][/a]

Ok, so according to Jonathan the idea is that it is a good thig to stretch the histogram to the endpoints if they are reasonably close (what does that mean? a gap of 20% of the width or less?)

What about underexposed images where the right hand side of the histogram is at, say, 3/4 the way across? Like this: |xxxxxxxxxxxxxxx_______|

Barring the "foggy scene" then, should we usually stretch the histogram to full scale?

[Tim Gray]

Barring the "foggy scene" then, should we usually stretch the histogram to full scale?
[a href=\"index.php?act=findpost&pid=146154\"][{POST_SNAPBACK}][/a]

Subject to having sufficient data/content, in general, yes - but it's for esthetic reasons, not technical.    

A better articulation of "best practice" is that you should ALWAYS know where your black and white points are and adjust according to your intent.

[Jonathan Wienke]

Ok, so according to Jonathan the idea is that it is a good thig to stretch the histogram to the endpoints if they are reasonably close (what does that mean? a gap of 20% of the width or less?)

I normally do a levels adjustment so that 0.01% of the highlights and shadows are clipped to white and black, except for some foggy scenes and image with really saturated colors. For foggy scenes, I set white and black to aesthetically pleasing values, and for saturated colors, I pull in the white and black points only as far as necessary to get saturated colors in the highlights and shadows to print well (usually less than 15 levels from min/max).

[marcmccalmont]

A daylight scene has up to 17 stops of light
Your eye at a given pupil size can see about 12 stops of light
Print film and MFDB's can record about 12 stops of light
A DSLR can record about 8 stops of light
Slide film can record about 6 stops of light
Your printer can print about 4 stops of light

So what every one is saying is, unless the original scene had less dynamic range than your printed paper is capable of (fog etc) the process is one of compressing the dynamic range into the maximum your printer/paper can display for best results.
Marc

[CUclimber]

Focus on getting good pictures, not good histograms.

I think that trying to apply a general rule to all histograms is the wrong thing to do.  I've taken plenty of fantastic shots that have "bad" histograms with clipped highlights, tons of 0,0,0 pixels, gaps, and everything else.  Not every picture needs to have a nice centered mountainous histogram shape.

[eonasick]

Focus on getting good pictures, not good histograms.

I think that trying to apply a general rule to all histograms is the wrong thing to do.  I've taken plenty of fantastic shots that have "bad" histograms with clipped highlights, tons of 0,0,0 pixels, gaps, and everything else.  Not every picture needs to have a nice centered mountainous histogram shape.
[a href=\"index.php?act=findpost&pid=146432\"][{POST_SNAPBACK}][/a]

Yes, it's a given that getting a good image overrides everything else. Canadian photographer Freeman Patterson has me jumping up and down, breaking rules, and all sorts of bad behaviour - with great results. As a friend of mine says: "It's taken me 20 years to get my pictures sharp; now it'll take another 20 to get them blurry...

However, it's good to know your tools, their purpose, and their limits.