Setting the Black and White Points

An essential function of preparing an image for display is that you literally have to tell the computer what is black and what is white. Often after a preview scan is performed, you will note a tone that should be black is lighter than black and a tone that should be white is darker than white. On the other hand, the computer merely knows that pixels with a value of zero should be shown as black and those with a value of 255 should be shown as white. As far as it is concerned, the preview scan is an undifferentiated mass of pixels; some may even have values zero or 255. This process of telling the computer what is black and what is white is called setting the black and white points. The importance of this matching process is less that the extremes are defined but that the entire continuum of tonal values for the image is defined. To be sure, most scanning programs have functions that automate this matching by finding the darkest tones and forcing them to black and forcing the lightest tones to white.  A skillful scanner operator, however, will be able to ensure that the correct selections are made and override the defaults if necessary.

fig. 1 fig. 2
Failure to specify the black/white points causes the computer to display black and white improperly (fig. 1).  An image with the black/white points properly set will have a histogram that spans the full range of tones (fig. 2).

In this example, the initial scanner histogram has gaps between the margins of the graph and the left-most and right-most portions the histogram so that the darkest tone in the image is not black and the lightest is not white.

Input histogram

Setting the black and white points may be accomplished in several ways. The most direct way is move the black pointer (b) until it is immediately below the left edge of the histogram and the white pointer (w) until it is immediately below the right edge of the histogram. Using the black eye dropper to sample an identifiably black tone and the white eye dropper to sample an identifiably white tone have exactly the same effect. Another method is to press the 'auto' or the 'cont' buttons and let the computer do it. NikonScan, however, doesn't merely place the pointers at the extreme edges of the histogram. For the black point it moves the pointer to the right beyond the left edge until the darkest 0.3% of the pixels become black. For the white point it moves the pointer to the left beyond the right edge until the brightest 0.3% of the pixels become white. (You may redefine the black and white point settings in the 'misc' color tab in the preferences.  Note that Photoshop sets its clipping point at 0.5% at both ends.) The reasoning behind this is that NikonScan assumes the extreme 0.3% of the pixels are unrepresentative of the image (e.g. noise) and can be thrown into the extreme values; they are 'clipped'. In addition, the way in which the auto b/w point is set changed with NikonScan 2.5. Instead of operating on the RGB tone curve (going top-down), NikonScan individually sets the b/w points on the underlying red, green, and blue curves (going bottom-up). Theoretically this achieves the greatest tonal separation.  If the b/w pointer settings among the red, green, and blue curves differ greatly, however, it means that the individual color channels have undergone radically different contrast adjustments.  It may be difficult to restore the color balance using only curves. The balance may be achieved by adjusting the analog gain of the underlying color channels until the relative positions of the b/w pointers among them are restored.

Curve with b/w point setting and output histogram

When you set the b/w points you're really making 2 curve adjustments: the left-end of the histogram is shifted down to zero, and the right-end is shifted up to 255. The net effect is a contrast change curve applied to the histogram. The equivalent tone curve method to setting the b/w point is demonstrated in fig. 1.  Note that the curve (in red) has an "S" shape characteristic of an increase in contrast.

Setting the Black-White Points in Hardware

In contrast to the previous software method using curves, setting the black-white point in hardware works in reverse by modifying the shape of the histogram until it fits within the graph's margins.  On the Nikon LS, for example, this means using analog gain, which controls the luminance of the scanner LEDs.  Increasing analog gain is similar to increasing the exposure in an enlarger so that more light passes through the denser areas of the image.  Usually this method is used in conjunction with transparency films, which are more likely to have dense tonal areas.  The reason for using this method is that by forcing the image to span the scanner's density range in the analog domain samples are obtained potentially for all 256 values, assuring a continuously toned image.

Here is an example using this method done with a Kodachrome image:

This is the initial preview scan with no analog gain.  The image utilizes approximately 60% of the scanner's density range.

With AG set to 1.0 the right edge of the histogram reaches 85% of the white point.

Setting the AG to 1.5 completes setting the white point.

This image was scanned by using curves only. Its histogram reflects the fewer samples made available by setting the black-white point in software.  Since this histogram is for the preview image and not the larger scan, it looks worse than it actually is to the hardware method, as can be confirmed by comparing the images.  In fact, because of the limitations imposed by this medium, the images are too small to discern any qualitative difference.

Here is an example where using analog gain results in more image shadow detail (turn up monitor brightness all the way):

The differences between these two images illustrate the magnitude of quality between an optimally and sub-optimally scanned image.

Close this window