Reading Histograms

[bjanes]

Check out MRs tutorial http://luminous-landscape.com/tutorials/expose-right.shtml which also contain a reference to another good link  "Understanding Histograms".
[{POST_SNAPBACK}][/a]

MR's essay is what introduced many of us to the advantages of exposing to the right (ETTR) and subsequent comments Thomas Knoll and Bruce Lindbloom were very helpful.

From my reading of MR's essay, he is relying on the histogram and flashing highlights to determine when highlight clipping occurs. For most work this is probably sufficient, but with most cameras clipping in the RAW channels does not occur until exposure is 0.5 to 1EV over that indicated by the above methods.

If you want to carry out ETTR fully, it is necessary to take a spot meter reading of the highlight and apply an exposure compensation that has been determined by previous testing to place the highlight just short of clipping in the RAW channels. This is often about +2.3EV over the metered reading. This is discussed at length in a thread on the Adobe ACR forum:

[a href=\"http://www.adobeforums.com/cgi-bin/[email protected]@.3bb6a85c.3bbb579c/141]http://www.adobeforums.com/cgi-bin/webx?13...5c.3bbb579c/141[/url]

Of course, if you miscalculate, you risk blowing the highlights and losing data. IMHO, if the shadow portion of the histogram is not clipped, the camera is capturing the full dynamic range of the scene. ETTR will give less shadow noise and more levels in the shadows, but this may not be worth possible loss of highlight detail. Moreover, most raw converters roll of the shadows anyway, Nikon Capture more than Adobe Camera Raw.

Bill Janes

[bjanes]

That's not quite accurate. The jpg has had contrast settings applied and has been converted from the raw data into a smaller color gamut, smaller tonal range and a different gamma than what is available from a raw file. The histogram will show clipping on the jpg before clipping would normally occur on a raw file.

To get a histogram that most accurately represents the raw file you should reduce saturation and contrast settings to 0 and set the color space to AdobeRGB. Even this is more constrained than raw data but it gets you as close to it as possible.
[a href=\"index.php?act=findpost&pid=57706\"][{POST_SNAPBACK}][/a]

Since the contrast merely applies an S curve to the data, the quarter tones are most affected, but the actual highlights are little affected and I don't think that the histogram would change that much.

Here is a graphical output of a Stouffer step wedge (0.1 density steps or 0.3EV) of Nikon Capture renderings at normal, low and high contrast. If all the grid lines are not visible, view at a larger size.

Bill Janes

[dwdallam]

Just by a light Sekonic meter and forget about the histogram, or bracket everything. Blah blah.