Modifying Brightness and Contrast
The properties of the tone curve (red, above) imply
Here is what we do
The end-points are absolute, 0 and 255, so the contrast between black and white cannot be increased. (In this tutorial, I imply two types of contrast: local and global. Local for contrast between two tones. For an entire image's contrast, since the end-points are fixed, I look at the histogram's standard deviation as an indicator.) We can, however, increase the contrast between two points if one is not lying on an extreme. Increasing the slope of the curve* between the 2 points to more than 45º has the effect of spreading an input tones over a wider interval of the output line.
When I say slight, I mean slight. To increase contrast in one range of tones means to decrease contrast in another. If you go too far, tones in the increased contrast area will start to lose their continuity and posterize (fewer tones are distributed into a larger number space), and tones in the decreased area will begin to merge (more tones are squeezed into a few values). We're restricted to working within a very small interval of 256 numbers--a zero-sum game. We can perform adjustments on an image, not miracles. Even the most expert scan cannot compensate for a lack of tonal range in the source image.
Keep in mind that each pixel in a color image is a composite of red, green, and blue values. When an RGB curve is applied to an image, the scanner software or image editing program applies the curve to the underlying components separately.
An interactive demonstration of curves (requires Java enabled browser)
Identifying Tonal Areas as Grab Points
Of course, when we modify a curve we don't grab just any point, we manipulate the curve with specific tonal areas in mind.
Changes in contrast requires that two points be identified. An increase in contrast is achieved by moving the two points so that the slope, or angle, of a line drawn through them is increased. A decrease in contrast is achieved by moving the two points so that the slope, or angle, of a line drawn through them is decreased.
With practice, knowing where to grab the tone curve becomes almost a natural reflex.
Ameliorating the Loss of Tonal Separation
There are three methods to address the distortions caused by the application of tone curves:
That's the theory and received wisdom. In practice, it's rare that methods 1 and 2, although resulting in better looking histograms, ever result in visually improved images. The one possible exception arises when scanning a transparency with important shadow detail or a transparency that is underexposed. If the source image is properly exposed and won't undergo extensive tonal modification, 8+-bit scanning is a bit of a waste. In any case, if you're determined to scan at greater bit depths be sure to use multi-sampling. It makes little sense to attempt to obtain the benefits of the precision of 8+-bit depth scanning without the increased accuracy of multi-sampling.
You might wonder what the role of the sliders might be in the method I have outlined. Simply speaking, none. I hate to sound preachy and dogmatic, but if there's anything I feel strongly about, it's that one should not use the sliders to adjust contrast, brightness, or color balance in creating a final scan. This is not especially because in NikonScan these adjustments are not reflected in the histogram, thus giving you no objective feedback, but because the effects of the sliders are so crude and thus work against the objectives of proper tonal correction. This may be demonstrated by analyzing the sliders as tone curves.
Naturally, this remonstrance applies equally to using the tone control sliders in imaging editing programs such as Photoshop.