I am currently working on an authentic digital version of the discontinued Kodak Color Infrared film (Aerochrome) with cameras that have a Bayer sensor. Of course, I’m not the first to do that.
Various IR fans have tried this, some obsessively, over the years, some with good results, some rather poor. The optimal solution does not exist today. The matter is difficult. But it would be half the fun if it were easy – then you’d probably find the look all over Instagram already, besides I’m learning a lot about how cameras work at the moment.
The main problem is that DSLR cameras capture IR light (or more precisely near-infrared light) differently than Kodak Color Infrared film. Simply put, the difference is that Kodak Color Infrared film “processes” NIR light separately, while DSLRs treat NIR light as part of red and also distribute it to green and blue. Aerochrome registers red, green and NIR, while DSLRs register red + NIR, green + NIR, blue + NIR.
William H. Klein explains the process in film in his 1980s paper Understanding Color Infrared Photography:
Color and color infrared film is comprised of three emulsion layers, superimposed on each other on a triacetate base. Each emulsion layer contains dye-forming silver halide compounds that are sensitive to certain portions of the photographic spectrum.” These emulsion layers are “sensitized by the green, red, and near-infrared components of the spectrum”, and “blue light is prevented from reaching the film by a yellow filter” placed over the camera lens.
Since infrared light, or more specifically, near-infrared light, forms a separate “color channel” in Kodak Color Infrared film, it does not contaminate the other channels.
DSLRs thus have two problems:
1. there is no filter (at least none for the mass market) that only transmits red, green and NIR.
2. even if there were a filter, the problem of NIR contamination of red and green would not be solved.
There is the unsatisfactory two-photo solution with full spectrum camera: You take two pictures (with tripod) of the same scene. One with a filter that allows only NIR light and no visible to pass (i.e. from 830nm). Another one with hot mirror, i.e. only with visible light.
Then you replace the red channel in the hot-mirror image with the infrared image and shift the (hot-mirror photo) channels red to green and green to blue. This results in a NIR-red-green photo, as with Kodak color infrared film. Ready to use, without IR contamination of the other channels.
But who wants that? Always shoot two photos in a row, precisely matched to each other, plus always screwing the filters on and off.
There is a potential solution:
As far as I know, there is a filter that fulfills number 1. But this costs between 500 and 1000 Swiss francs, depending on the size. The IR contamination could be corrected in Photoshop afterwards.
And that would be relatively simple: As described in the first post on this topic, the color channels are swapped after shooting and white balancing, so that channel blue is shifted to red, red to green and green to blue. Blue then serves as the NIR channel. This would be very similar to the process used in color infrared film and very easy to accomplish.
To solve the contamination problem, the remaining blue (i.e. NIR) in the other two channels would then be subtracted in the channel mixer. This would probably work with a simple input there (per channel) of -100 percent for blue.
But does this really work? To my knowledge, no one has tested this. To my knowledge, there is no one who has bought this filter and tried this.
And an even more important question: Does it really make a big difference, respectively does Kodak Color Infrared film really have such clear properties?
Or to put it another way: Does the transmission curve of the Midwest Optical filter match Aerochrome exactly or are spectra excluded that are decisive for the look of Kodak Color Infrared film?
Is the solution with the orange-green filter (all the images shown in this post were made with this combo and minimal post-processing), which excludes blue and strongly reduces the incidence of red and NIR, possibly sufficient or even better to obtain a result optically comparable to Aerochrome?
Do we achieve all the Aerochrome typical color changes with this? Do we have enough variance in the colors of the different plants? And how much post-processing is needed to achieve this? And what are the concrete visual differences between the Orange-Green Method and the Two-Photo Method?
This is what we have to investigate now. Perhaps meaningful tests are possible only in the spring, when the plants bloom again.
