I just wanted to share something I thought about. I have not even Googled to see if there is a process out there for it but knowing how people tend to have parallel thought patterns, it would not surprise me if there are some other articles. In an article I posted quite some time ago, I talked about the Nikon Coolscan V film scanner that I had picked up. I was relatively pleased with the quality of the scanning but secretly have desired to see a more faithful reproduction of the original’s color and tonality.
How scanning works…
Thinking about the scanning process one can see that it is a rather difficult process. A lens is used to focus light of some sort that is transmitted through the film onto a sensor. In essence this emulates the process a digital camera uses to capture a scene with two major exceptions. First, the light of a scene is generally reflected from objects and rarely is transmitted through the target. Second, the image that is being scanned is very small relative to a scene in the real world — this means that there must be a high number of pixels per inch of the target compared to the real world scene.
This thought process should lead one to the conclusion that a scanner has a difficult life. It is doing all that it can to be as faithful as possible to the original image. Film processes generally have a much larger latitude than digital processes. Michael Freeman in his book “Mastering HDR Photography” pointed out that fine-grained panchromatic film can capture 4 orders of magnitude of brightness (10e4 or 1:10000) whereas current (as of the date of the book) digital sensors could capture less than 3 orders of magnitude (10e3 or 1:1000) (see page 15). With this being said, it makes sense that film scanners which use the same types of digital sensors are hamstrung by the same problems!
Using HDR to improve scan depth
What I decided to try was to attempt a scan on an image with three different exposure values. These could then be fed into a decent HDR program, reconciled into a 32-bit image, then tonemapping algorithms could be applied to produce a deeper image than might be possible with a single scan. Maybe, just maybe, scans could start approximating the true depth of detail available in the original.
A key requirement for this process is to ensure that the scanner software would not change the positioning of the image or any cropping dimensions between scans. Another requirement, of course, is that the software would permit exposure values to be varied across a continuum of several EV values around center.
I had the perfect test photo. There was this image I had taken on Ektachrome almost 15 years ago. I had stood on tip-toe at the base of a rock in the North Carolina Zoo’s RJR Aviary and captured a gorgeous orchid surrounded by ferns. It had a wide dynamic range and colors that ranged from delicate pinks to deep greens. There was nothing to lose!
Scanning and processing with Photomatix
The first step was to prepare the three scans. I opted to create three scans at center (0EV), -2EV underexposed, and +2EV overexposed settings. I turned off all sharpening and kept scratch and dust settings on their lowest values. This produced the three images below:
Once these had been acquired with exactly the same dimensions/crop, they were saved off then pulled into Photomatix Pro 4. The software created the 32-bit merge of the three images without any trouble and even recognized the EV steps between the separate images without my having to manually enter this data.
The 32-bit HDR file was then taken through the tonemapping process. The beauty of this is that this process permits quite a bit of control. In fact, many different permutations of the image could be generated.
I won’t get into the details of my processing in Photomatix here. If you are interested in knowing more about the tonemapping process, let me direct you to look at part 4 of my HDR series (see My Adventures with HDR: Part 4). You can follow the links back through each of the parts.
The tonemapping process then produced a 16-bit TIFF image. This is what it looked like:
I believe that if you compare this with the 0 EV scanned image above, you will see that there is a lot more depth to the flowers even in these small versions!
Final processing
I am one of those people who delight in tweaking a final image in Photoshop. Regardless of if I am working on HDRs (with Photomatix) or Panoramics (with Autopano), the image I prepare in these packages is merely meant to trampoline me into my Photoshop workflow. Thus, I sucked in the TIFF output from my tonemapping exercise and applied some final color balancing, saturation/desaturation, tweaking of curves, and selective sharpening to produce a final image. It was indeed a pleasure working with this image since it did have quite a lot of latitude. Below is a small view of the final image:
A larger view can be seen on my Flickr photostream at www.flickr.com/photos/claforet. It is plain to see that the process did work rather well. While I don’t advocate this extended workflow on every scanned image, there are definitely those which merit such painstaking attention. The reward is well worth the extra effort that it takes. I cannot wait to try it on some B+W negatives to see how the dynamic range stacks up!
As always, I am open to comments and suggestions.
I recently took a pair of really old black and white negatives and worked them through this process also. I was relatively pleased with the results. There is a good dynamic range in the resulting digital image. See http://www.flickr.com/photos/claforet/5670891773 and http://www.flickr.com/photos/claforet/5668072484 to see the results and discussion.