Highlight Recovery

This Article is now obsolete due to the introduction of the new Recovery Slider in Camera RAW 4.

Anyone with a decent digital camera will be aware of the way the camera warns of overexposure by flashing the highlights on the LCD display. This is very useful and is a great aid to using the ‘Expose to the Right’ (ETTL) technique used by most knowledgeable digital photographers when shooting in RAW mode.

ETTL involved exposing the shot such that the shows the brightest highlights being as far to the right as possible without them being clipped off. This maximizes the quality of the data captured for later conversion from RAW.

Whilst this is quite simple to use, there is a slight fly in the ointment because the histogram display is based on a Jpeg preview created by the camera. In other words the histogram does not show the exact RAW data histogram but instead shows a histogram based on the Jpeg that the camera would produce if it was set to Jpeg mode. The upshot if this is that the little histogram on the back of the camera cannot be completely trusted and should be treated as a guide only.

The other limitation on the camera histogram and the flashing display is that it shows clipping of any RGB channel, not all of them combined. Thus if only one colour channel is clipped the display will flash and you will think you have overexposed the shot. Actually, you may not have, maybe only one channel is clipped with the other two being perfectly OK.

The reason that this is important is that Adobe Camera RAW, as well as some of the other RAW Converters, can interpolate detail out of the unclipped channels and recreate some information for the clipped channel – this is known as Highlight Recovery and can be surprisingly effective. In fact it is possible to ‘recover’ as much as two thirds of a stop if only one channel is clipped and one third if two channels are clipped.

Here is a shot of a Buddhist monk I took on a recent trip to Cambodia. The light coming through the wat shutters is quite strong and the background is rather dark. When I looked at the camera display I saw that some of the monk’s face was clipped and small flashing highlights covered his face.

On opening the file into ACR I checked the Highlight warning box and saw this:

On my usual defaults there was very little detail in the highlights on the monk’s face and had I been shooting Jpegs this would have been almost impossible to correct. However, in ACR I have two powerful tools at my disposal – the combination of Brightness and Exposure plus the Curves tool make it fairly straightforward to bring full detail back into the bright highlights in this image.

Brightness and Exposure

The Exposure slider shifts the overall histogram left or right and so by setting it to minus one I have corrected those clipped highlights and brought all three channels back into control with none clipped – the red overexposure overlay has now vanished indicating that no channels are clipped. ACR has rebuilt the clipped channel and detail is clearly visible where before was only featureless white which would have printed very badly.

However, by reducing the exposure by one full stop, the image is now clearly darker and this must be compensated for by using the Brightness slider to bring back the overall luminosity of the image to satisfactory levels. A setting of 90, up from my default of 70, seems to do the trick. Note that using the Brightness slider does not push the overall histogram up, it shifts the shape of the histogram so that the midtones are now where they were before we reduced the exposure. In effect we have pulled the highlights back into line whilst at the same time keeping the all important midtones much the same.

Curves

ACR has a powerful tool under the Curves tab, much like normal Photoshop curves, but these curves act on the non-linear RAW data before a gamma correction is applied. Put simply, these curves allow very subtle adjustments to be made to what might look like flat highlight areas but in fact can contain as much as half the tonal detail of the entire image. By steepening the slope of the curve in the highlight region better separation can be revealed between highlight tones that would otherwise be lost when the image is processed into a normal TIFF image.

In this example I have changed the curve very slightly from the default and, whilst it may not show up here on the printed page, this adjustment has revealed a great deal of tonal detail in the very lightest of the skin highlights.

Used with care and a light touch this tool can make a great deal of difference to most images, especially ones with important highlights.

Nick Rains 2006