Tri-Chromic Narrow-Band Lighting for Cam-Scan
Tri-Chromic Narrow-Band Lighting for Cam-Scan
On another forum, people are experimenting with tri-chromic narrow-band illumination for camera-scanning of color negatives. Explanation: It's not continuous spectrum lighting (like sunlight, incandescent, or electronic flash, what we see as white is a combination of ALL visible wavelengths). Instead we are talking about pure red, pure green, and pure blue, nothing in between (with the right mix, we see it as white, our RGB monitors work this way).
How to get such light? A current iPad or iPhone is pretty good. Stage/theater lighting with R, G, and B LEDs (leave the white ones off). For $30 at Amazon, Sansi offers a pretty good RGB yard light. OK, let's try.
There are two approaches:
- One-shot, with all three light colors on.
- Three shots, one w/Red, one w/Green, one w/Blue light only.
First, here's my current reference shot, Fuji 200 film, a new shot, full fall NE sun. Sony RAW cam-scan for ETTR at +1.5EV with electronic flash. Converted in LR with Negative Lab Pro. Settings: Frontier 3 Std Auto-Color Neutral.
This is pretty good. I would adjust the hues for Kodak yellow and Coke red, but I think this is a very workable scan. What will tri-chromic narrow-band lighting do with this negative?
One-Shot Narrow-Band
Same method. Sony RAW file, negative illuminated by an iPadPro with screen at max brightness, white light. I don't know whether camera engineers ever considered such lighting. Same processing as above.
Result is pretty good with these setting and no adjustments. Not as good as above. Kodak and Coke have drifted a bit further from desired, but all fixable. My conclusion: lights like these are fine for camera-scanning. Many people are using iPad, phone, or computer screen for backlight. Drawback is the ~2-3sec exposure. Here's the result:
Three-Shots, R G & B, Combined
This starts to get crazy. What to do with the three shots?
Some have built a "Frankenstein's Monster" image file by extracting one channel from each of the three images, then somehow processing that. I found mixed results from these, and was ready to abandon the whole thing.
Then an interesting suggestion:
- Run the raw files through MakeTiff, which I own, to de-mosaic and turn into linear tiff files
- Load these into layers in PS, aligning the layers
- Set the layer blend mode for all layers to "Lighten"; this produces a viable negative image.
- Levels layer to adjust histograms for all three channels, normalize each channel, full width, no clip. Leave middle slider at 1.00.
- Curves layer to invert with a big curve (I used a big bend in the curve, see below)
- Curves option-Auto, snap neutral midtones, for a basic auto correct.
Sounded crazy to me, but here's the result: Lots of good rich color here? Coke is good, E100 is good, Kodak yellow quite good. Gray stones and black t-shirt are better than either of the above conversions.
You can
download the inversion curve, or make one that looks like this:
Some of my pals are using this method as their standard, creating automated tools to take the three shots, load and convert. The question is whether it's worth it. Does the three-shot capture give more color separate and better color? It was used in CoolScan scanners. Was that for quality? stablility of light source? or to make scanners less expensive? Anybody know?
Later I'll add links to the files in case anyone wants to try this.
Comments welcome