Hi,
First the distinction between raw and b/w histogram. The raw histogram represents the pixel data as they come from the scanner, so whatever you do *after* scanning, won't have an influence on those. The b/w histogram represents the pixel data in the final, postprocessed image. Both are quite different, for one thing, the raw histogram is for the negative image (thus, low value = shadow, high value = higlight), while the b/w histogram is for the positive image (low value = shadow; high value = highlight).
Some background on the raw histogram. I'm not 100% sure, but I assume that the axis of the histogram shows the (base 2) logarithm of the pixel values. E.g. if you scan 16 bit, the pixel value logarithm will range from 0 to 16.A scanner is a linear device, this means that the amount of light transmitted by the negative has a linear relation to the pixel value. If some pixel A is twice the value of the other pixel B, this means that at that spot, the negative transmitted twice as much light than at the other spot. In logarithmic terms, A-B=1.The density of the negative on it's turn, is linear to the (base 10) logarithm of the transmitted light.
Putting this together, the values on the axis of the histogram are linear to the negative density.
This is useful to explain some behaviours:
- the density range of the negative is not so much an effect of the (film) exposure, but rather of the development, so it doesn't surprise that your overexposed negative doesn't show a different density range
- if you change the scanner exposure, the histogram will not become wider or narrower, but it will shift along the logarithmic axis.
- a negative has a maximum density range of ~2.5 density steps (base 10 logarithm values), which translates into ~8 steps on the histogram axis (2.5 * log(10)/log(2)). That's the maximum width your raw histogram wil ever get. Underdeveloped films have lower density and will have a narrower histogram
So we have two parameters to control the histogram:
1) development will make it wider or narrower
2) scanner exposure will define the position on the histogram axis
What's the optimal tuning?
Let's look at the highlight end of the axis (that's the low pixel values, because we're talking negatives!). There are two quality threaths here.
First there's the scanner noise. Because the scanner is digitizing low light values, the detector noise will have a high impact on the signal.
Second, there's the low 'dynamic range' at this end. Consider a high density gradient area on the film ranging from density D to E, spanning a difference of .3 in density steps. On the logarithmix pixel value axis, this translates into a difference of 1, meaning that pixel value D is twice pixel value of E. If the histogram is positioned to the low end of the pixel data range, values D and E could be e.g. 8 and 4. There's only 4 pixel values inbetween to capture the gradient on the film. If the histogram is shifted 5 steps to high end, values would be 256 and 128. That gives a much smoother gradient!
Conclusion is that the low value area of raw pixel data should be avoided.
This is achieved by
1) avoiding high density range on the film to keep the histogram relativelynarrow
2) targeting at the highest scanner exposure, to 'bump' the histogram as far as possible into the high pixel area, without clipping.
Groeten,
Vic
