For a while I have got developed and scanned my 35mm negatives by a lab/shop. The quality of the scans is not satisfying me (a lot of noise and images are not sharp). I was thinking of buying a film scanner (thought of plustek 8200 film scanner). But, since one of the weak points of these scanners is usually the dynamic range, I wanted to know if it would be possible for me to scan the same frame multiple times (2 or 3, much like an HDR) with different exposures (since I understand you can adjust exposure of the scan through the software) and then merge it through lightroom into an image with more dynamic range, in order to get closer to the original DR of the negative
Simply put: No. You don't really need to worry about dynamic range if you are planning to scan ordinary color negatives. A scanner's Dmax becomes a limiting factor only when scanning high-density transparency material like color-reversal films. In order to tell you why I have to clarify a couple of concepts.
- Film Density
As you expose and develop a film, different parts of the film surface become transparent to different degrees as a function of the light exposure and the developer they have received. The density of the film at each point refers to the fraction of the light that gets through the film if you illuminate it with some light source.
Please note that film's density is only a translation of the original contrast of the photographed scene. Different emulsions can develop different densities in response to a light exposure of the same intensity. The good news is that negatives compress the contrast of the photographed scene very effectively in a pretty narrow density range. Color reversal films, on the other hand, produce a very high density range while they are not as good in compressing contrast.
- Scanner Dmax
A scanner's Dmax is the ratio of the strongest light signal to the weakest light signal it can record, and it varies with the sensor's bit depth. As you can imagine, the higher the density range of the film you are scanning, the higher is the Dmax you need to represent that density range in digital space.
Note that while scanning, your scanner doesn't deal with the original contrast ratio of the photographed scene, but rather with the density range of the transparency it is scanning. So if your scanner has limited Dmax, it may still be able to represent the information contained in your film pretty well if your film's density range is limited. This is actually the case for ordinary color negatives. Good color negatives are known to record a dynamic range of 14-15 stops (a.k.a high exposure latitude) but fortunately they compress all this information in a narrow density range.
The result is that even a mediocre scanner with a working Dmax of 2 must be able to make a decent scan of any color negative. Most of today's negative-capable scanners can easily handle this. Very contrasty B&W negatives may have a slightly higher density range, but in my experience they are usually within the capabilities of the consumer-grade negative scanners.
The slides are very different though. The image they record is very contrasty: They can record a small fraction of the light intensity variations of the original scene (4-5 stops maybe - a.k.a low exposure latitude) and that over a very high density range (up to 4.0). That is where consumer scanners fall short of scanning film and the software solutions like multi-exposure are developed to deal with the physical Dmax limitations of the scanner sensors in such cases.
So, if your target media is color or B&W negatives, go ahead and buy the scanner that fits within your budget and delivers enough effective resolution for your needs without worrying much about the dynamic range. Your scanning technique will ultimately count much more than your scanner. But if you are planning to scan a lot of slide film, you may want to hold back until you can buy a top-of-the-range scanner like Epson V700/800, Plustek 120 or Nikons. Apart from their Dmax, these scanners can be very accurately color-calibrated, what is in my opinion crucial for scanning slides.
Making HDR images out of multiple scans of a slide film at different exposure levels is a possibility, but is not a practical solution to Dmax limitations. The color response curve of slide films is not homogenous across the range, and the IT8 targets you receive with top-of-the-line scanners are supposed to help you obtain natural-looking colors from your scans (IT8 target is a must for any color slide scan). Exposure variations mess up the color corrections brought by the calibration algorithm and end up producing faux colors. That is one of the reasons why the multi-exposure option of the softwares do not perform very well: If the exposures are too close the added contrast information is not significant, and if the exposures are too varied the color balance goes wrong. You can probably do a bit better than the automated software if you process the HDR image manually, but in my experience the outcome is not satisfactory and not really worth the time and effort.
For more technical reading on dynamic range and Dmax: http://photo.net/learn/drange/
Sometimes one image is better than a thousand words. The samples below illustrate very well what I just explained above. The scans below were first saved in 16-bit tiff format and then converted into 8-bit jpeg manually.
- A very high-contrast scene shot on Kodak Kodak Tri-x 400 (pulled by two stops). The photo was scanned as a raw negative file (linear tiff) using the open source SANE Pixma backend on the affordable Canoscan 9000f mkII. The subject is in direct sunlight and there are deep shadows behind her in the trees. Note that the image represents the unmodified linear scan (Gamma 1.0), which is the reason why it looks so dark.
- The scanner has recorded the full dynamic range of the film with little effort. The density range of the negative film seems to be lower than what the scanner's Dmax can cover: The corners of the histogram contain little information.
- The same image after gamma correction. You can see that all the details of the bride's white dress are captured. The deep shadows in tree leaves also seem to contain plenty of details.
- Image histogram after gamma correction.
- Inverted positive of the same Tri-x shot. The image has been contrast-adjusted after inversion. I had to clip some details in the dark areas deliberately in order to obtain deep shadows and a 3D pop.
- The histogram is still pretty smooth after contrast-stretch. Converting the image to 8-bit will give you a round and curvy histogram with perfect lightness gradation (no dents in the histogram).
- A medium-contrast scene shot on Fuji Velvia 50 (highest film density) and scanned with a V700 on a color-calibrated copy of SilverFast. No multi-exposure or other digital techniques have been implemented to widen the dynamic range and no contrast or color adjustment has been applied (Except the scanner's default Gamma correction and IT8-based color correction). In reality the scene looked much less contrasty than what you see in the picture. The sky was covered with clouds and all the shadows were soft. Yet, as you can see, the sky looks washed out.
- A look at the image histogram tells you that the slide's density range has covered the full dynamic range of the scanner. The scanner hasn't fallen short of scanning the slide awfully, although there is some clipping in both highlights and shadows.
- The same slide scanned with the affordable Canoscan 9000f mkII (uncalibrated) and the open source SANE Pixma backend. I scanned the image as a linear tiff, and then did the color conversion using ColorPerfect. The raw file is way too dark and useless to post here, but the final result is acceptable. Note the bizarre red/magenta color cast and the unnaturally vivid colors. IT8 helps eliminate these. The colors in the scan above look much closer to the original slide when put side by side.
- The histogram is pretty acceptable, although we can see there are signs of clipping before color conversion. Also we can see a lot of irregularities in the highlights across color channels due to lack of calibration. But overall, the details in dark areas hold up pretty well against the output from V700, what I find quite surprising.
I believe this feature is already built in to some scanning software; you don't even need to do it yourself manually. There are two approaches, for slightly different outcomes:
Multi Exposure scans the film multiple times using different exposure durations, and combines the images to simulate extended density range (while also reducing grain).
Multi Sampling (which might be called Multi Exposure in VueScan?) scans the film multiple times with the same exposure time, and averages the readings to reduce noise, get more natural colours and achieve smoother colour transitions.
A side note - Nikon's scanner software doesn't allow these operations with the Coolscan V film scanner, even though the hardware is capable of performing them. (Third-party scanner software doesn't have this limitation with this hardware.) I guess Nikon wanted to increase differentiation with the Super Coolscan 5000 model.
My info is from the book 'Scanning Negatives and Slides' by Sascha Steinhoff.