A few scanning tips


What about copying color negatives with a DSLR?

Color negatives are a special problem, to remove the orange mask. Film scanners scan color negatives with a longer exposure time of the blue and green channels, perhaps 4x longer (than red) for blue, and perhaps 2x longer for green. This acts as an analog glass filter at the lens, and the longer exposure boosts the blue and green, leaving the orange complement behind. Otherwise, when inverted, the color negative film appears deep blue-green (complement of orange) instead of correct color.

Some choices:

Adobe AutoLevels almost works then, but is not totally satisfying. It is a rather difficult job for color negatives. B&W negatives are easy, you simply invert them and you are done. But you really want to use a film scanner for color negatives if possible.

Or, if you have VueScan, it will "scan" camera images from a JPG file (Source: File, and Media: Color negative), and convert them to positive color. Pretty neat.

Less satisfactory than any IMO, you can select the middle Gray eyedropper in Levels (or Photoshop Curves) and click some area that should be white or neutral gray, and it will convert the entire image, removing any color cast there. You could scan a piece of clear blank film leader that is orange, and use that instead (either inverted or working on the uninverted image). One problem is that the intensity of the orange mask varies, it has lighter and darker areas, it is not a constant everywhere, and your film leader may not be the best sample.

Regarding color negatives, this next manual procedure almost works too, in very much the same way as AutoLevels. It is presented mostly for the concepts involved, rather than a working tool. But it should be much more clear after trying it once.

Red Blue Green RGB Final

After inversion in the photo editor, you see the image with a deep blue bias (the complementary colors), instead of the original orange mask. When you examine that histogram in the Levels tool, you see that the Red channel data pretty much begins near zero, like any regular image (shown, but leave it alone right now). It is slightly clipped here (the little spike right at zero).

But the Blue and Green channels are shifted much higher, with the data not beginning near zero (due to the orange mask). The Blue data begins maybe about 100 here (exact value depends on the exposure of the frame copy or "scan"). The Green data begins nearly as high, perhaps 80.

To correct this blue-green bias, move the Black Point of the individual Blue and Green channels up to where the bulk of the data noticeably begins (as shown, ignoring any little tail of zero height). This sets the new zero point to be where you set it. The color seen in the Preview image is almost corrected.

This is very similar to the film scanner exposing these channels longer, moving them higher (brighter), which causes them to be lower (darker) after inversion. Inversion merely flips the histogram, so that 255 becomes 0, and 0 becomes 255.

Perhaps some fine repositioning of the Black Points can do minor color tweaking. You can see the resulting color in the Preview image. This is about what the AutoLevels tool will do too. Typically, AutoLevels clips both ends of individual channels by about 0.5%. I am clipping them a bit more here, due to cutting of the thin trailing tail (very optional however - experiment with both ways. It does affect color and contrast).

The idea of the rough adjustment is to make the blue and green channels also begin near 0, same as the red (to remove the blue cast). This is about what the film scanners extra exposure of the blue and green channels would do too, boosting blue and green to offset the orange, which is a lowering operation after inversion.

Note: These corrections are based on the data "that is", whatever is in the histogram. It is NOT about the original image, or how the image ought to look. White Balance attempts have no clue what the image is, or how it ought to look.

Then select the Composite RGB channel to brighten it if necessary. The idea of using the composite RGB channel (instead of individual channels) here is to avoid affecting color with this adjustment. Anything done differently in the individual channels affects color balance. But the same thing to all channels normally affects brightness and contrast, but not normally color balance. Move the RGB White Point down to where the bulk of data curve begins (as shown, ignoring any little tail of zero height). Then press the OK button for all of the operations above.

The final result is shown last. Converting to 16 bit color first will smooth the histogram some, but won't have visible affect on the image.

In Adobe Elements, the menus are not quite obvious. Select the image, and then Invert is CTRL I, or menu Filters - Adjustments - Invert. Levels is CTRL L, or menu Enhance - Adjust Lighting - Levels. Or, in the Levels box, you can select the Black Eyedropper, and then click the darkest area of the image that you want to be black, and it will set the Black points there, in a way very similar as described above.

More about histograms and Black Point/White point is shown elsewhere here.

This is only about the general idea. Best results will be more difficult than implied here. It is not a simple problem. The film scanners really do this best. And for another thing, the original image may need some correction too.

Copyright © 2007-2013 by Wayne Fulton - All rights are reserved.

Previous Main Next