Camera Scan vs Film Scanner

The camera scan setup

My setup is economic enough to be a viable solution for ordinary quality-minded people (we are not targeting professional high-end labs here). The camera is a Canon EOS 7D, which has an APS-C size 18 Mpx sensor. The lens is a macro lens: EF-S 60/2.8 USM Macro.

EOS 7D was introduced in 2009, and a successor, Mark II, was introduced in 2014, and a Mark III is rumored to be launched in 2019. I might upgrade, and if so, I will make a new test to see if there are any significant differences for this particular usage.

I use an old slide duplicating attachment, originally designed for a bellows system, that I modified and adapted for this use. As illumination, I have a standard Canon flash that communicates with the camera to provide just the right amount of flash power for the current slide – i.e it adapts power depending on bright/dark slides. This flash targets a mirror in front of the setup, angled so that the flash bounces back right onto the slide duplication film holder with its opal diffusing glass. I use a regular laser pointer to find the right distance and angle to achieve the right angle of the mirror.

The camera is set to 1/250, f/8, ISO 100, One-shot AF, with Zone AF, where I choose an area in the slide that is important with good contrast. I do not want to use a glass frame to keep the film completely flat, since glass frames are basically impossible to keep absolutely clean on all 4 sides with no micro dust, no fog, no newton rings etcetera. But a glassless mounted slide may bulge slightly, so I need an adequate depth of field. Tests show that I need to stop down to about f/8 to provide adequate depth of field throughout the entire slide. f/8 corresponds to a depth of field of about half a mm if the acceptable circle of diffusion is set to about 2 pixels. My own tests of EF-S 60/2.8 USM Macro show that f/5.6 gives optimal sharpness in the center, whereas f/8 is the optimum in the worst-case corner. The difference in center sharpness between f/5.6 and f/8, however, is extremely small, hardly detectable at all, so diffraction is not a problem at f/8.

The slides are either mounted in glassless frames by me, or were delivered in plastic or paper frames from the lab. Before each scan, the slide is meticulously dedusted while using a good strong loupe. (You must use a loupe, otherwise you will be terrified when you later look at the image on your monitor in 1:1.)

The resulting RAW files, about 22–24 MB each, are imported to Lightroom. I have meticulously developed a so called Preset in Lightroom where I have adapted White Balance, Tone Curve, Shadows and Color adjustments to make the images on the monitor match as closely as possible the appearance of the slides on my Light table that uses 5-band 5000 K fluorescent tubes. (The monitor is profiled/calibrated to a white point that matches the light table reasonably well.)

The end result is an extremely close match between the images on the monitor and the appearance of the slides on the light table! And this holds for both Fujichrome Sensia and Kodachrome films! Even the perception of the grains is somewhat similar!

Since the sensor in EOS 7D has 5184 pixels horizontally, that equates to 3657 ppi (pixels per inch) for a 24 x 36 mm slide. That is slightly less than the 4000 ppi film scanner I will compare it with.

The film scanner I compare with

The film scanner used here to compare with the camera scan is a a Microtek ArtixScan 4000tf, and the scanning software is SilverFast Ai 6 from LaserSoft Imaging (http://www.SilverFast.de). A color management profile was meticulously created by inCamera Photoshop plugin using an IT8.7 reference target made by http://www.targets.coloraid.de, and with carefully optimized lamp settings for the three color channels to maximize the dynamic range. This way you can say that the scanner was carefully "calibrated".

This film scanner is now considered old, it was introduced in April 2002 and I bought it in December 2002 for about EUR 1135 (excluding VAT, converted from SEK to EUR). At that time, this scanner was considered one of the best CCD-based film scanners available in the market, with a dynamic range of of 4.3 stops and a resolution of 4000 pixels per inch, and capable of multisampling. It's primary competitor was Nikon Super Coolscan 4000 ED.

Although film scanners may have improved since then, the difference is not dramatic at all – the dynamic range is still about the same whereas some have higher resolution. But resolution is not at all important, whereas the dynamic range is the crucial quality merit that determines the capability to capture shadow details in slides. Also, the capability to keep the slide in focus is paramount. Some scanners use a directed sharp light source that makes microscopic dust much more visible, whereas a diffuse light source is more dust "friendly". Some have an extra infrared channel used to detect dust spots and mask them (replace with surrounding color); however, this technique does not work with Kodachrome films, and it will also degrade sharpness.

Despite ArtixScan 4000tf was considered a high end professional scanner, it had all sorts of problems, and it was a challenge to get perfect results. It's primary competitor Nikon Super Coolscan 4000ED had all sorts of other problems, such as jagged scans, almost zero depth of field that required the slide to be mounted in a glass frame. A look at the current 2018 offerings shows that they are as slow as the ArtixScan 4000tf at a similar price and with a similar dynamic range.

To really get perfect scans you would still have to use a dedicated drum scanner, but no private mortals can afford those, and scanning a large number of slides with such a monster would be a nightmare, including its use of greasy oil or vaselin on the film to mount it on the drum, and a subsequent cleaning of the film.

Comparing scans of a problematic photo

All images are clickable and open in popup-windows with (if needed) scrollbars. Inside each window there is text at the bottom saying what it is and its resolution. This way, you can compare different images by switching windows or placing them side-by-side. Some are downsampled, some are 1:1 images, i.e 100% pixel zoom. The images in the popup windows do not scale to your monitor – this means you can rest assured that the image size of these images are exactly as I have set them. All of them have been converted to the sRGB color space on export for this web page.

The two images below are a camera scan and an ArtixScan respectively of a Sensia 100 slide, shot with a Canon EOS 5 with an EF 20-35/2.8L. That slide is chosen as a critical test because it is an unusually difficult one: It has specular highlights, normal highlights of well known gradations, well known texture in water and buildings, and most importantly: it has details way down in very deep shadows. (The photo was taken in 2003 in Istanbul from the Galata Tower.)

The top image is the camera scan obtained as described above under section The camera scan setup. It matches the slide extremely well in all aspects as viewed on the light table! Even the subjective perception of grains is somewhat similar!

The bottom image is scanned by ArtixScan in SilverFast 6.5 with 8 times multisampling at gamma 2.0 (but unfortunately it lost sharpness in that process) as raw as possible ('48 Bit HDR Color'). Then my best color management profile for this scanner was applied (based on 8x sampling of an IT8.7 reference slide from coloraid.de). Thereafter I have lifted shadows 2% in Photoshop, then converted it to Adobe RGB color space. Then imported to Lightroom as a 113 MB zipped tiff file. See start of a tedious article series for further details on how I struggled to squeeze the most out of this scanner.

8-times multisampling is a bit extreme, but this is to see how much ArtixScan can push down noise, and how well a camera scan can compete with it. After import to Lightroom, it has had various tweaks to make it match the camera scan as close as possible so as to make it possible to objectively compare them: It has had adjustments for Exposure, Color Temperature, Tint, Shadows, Blacks and various fiddling in Tone Curve in all channels.

The result is that the ArtixScan is good, but it can not quite match the camera scan, no matter how much I fiddle with Curves. The difference is mostly seen in mid tones and in shadows. If you switch back and forth between the windows, you will see that they are not geometrically identical! Look at the buildings! It looks strange. The reason is that the step motor in the scanner is not perfect, so the distances it travels are not equidistant! The correct one is the camera scan.

The red rectangle imposed on the bottom image shows a detail of the picture that we will now take a closer look at in 1:1, i.e 100% pixel zoom.

Study of 1:1 details way down in deep shadows

When you click on any one of the three images below, it will open in a popup-window with (if needed) scrollbars. Inside the window there is text at the bottom saying what it is. They are not downsampled, i.e they are 1:1 images, 100% pixel zoom. The images in the popup windows do not scale to your monitor – rest assured that the image size of these images really are 1:1 pixelwise, and they are all converted to the sRGB color space on export for this web page.

The top one is the camera scan, i.e a 1:1 crop of the image shown in the previous section. It matches the slide extremely well as viewed on the light table! It has had my standard sharpening (50) and noise reduction (25) I always use for 7D at ISO 100. Even the subjective perception of the grains is somewhat similar. It does, however, have a slight tendency to banding in the very darkest portions. But I am pretty sure that the Mark II version of 7D would do better in this respect. Here is the same image without sharpening and noise reduction (but I think that all RAW images need some sharpening and noise reduction).

The next one is a 1:1 crop of the ArtixScan image that was shown in the previous section, i.e with 8 times multisampling at gamma 2.0. Multisampling often blurs the image due to imperfect mechanics, so in this case it unfortunately lost significant sharpness. What you can judge from this image is how much it suppresses noise and brings out structures in very deep shadows. You simply cannot get any better than this from the ArtixScan in terms of noise suppression and deep shadow performance (if we ignore pixel sharpness that was lost).

The third one is a 1:1 detail of a scan in ArtixScan using no multisampling at all. It has had some sharpening (25) and noise reduction (15). What you can judge from this image is how much sharpness you can get when you are lucky enough to get it perfect. It is the best of 4 scannings! You simply cannot get any better than this from the ArtixScan in terms of sharpness. Here is the same image without sharpening and noise reduction, which makes it significantly sharper at the cost of a disturbingly high noise.

Start by cranking up brightness on your monitor. Then click on the first two images to compare the camera scan with the 8 times multisampled ArtixScan by switching windows back and forth. If needed, scroll them up so you can see the descriptive text. Look at the lower part of the building in the middle, which has vertical stripes and very dark windows. Despite the multisampling in ArtixScan suppresses noise to bring out cleaner details in deep shadows, the camera scan brings out more details way down in these deep shadows! The difference may not be dramatic, but the camera scan beats the multisampled ArtixScan which becomes murky sooner than the camera scan.

Try the zoom-out function in your browser to study the images in less than 1:1 so as to judge the general perception of grains and deep shadow performance. Again, the camera scan wins! It looks cleaner with better blacks.

Now, close the ArtixScan, zoom back to 100% ("actual size"), and click on the third image to bring up the window of the single scanned ArtixScan. I think both sharpness and noise are similar to the camera scan, but the camera scan reveals much more of structures down in the deepest shadows. Maybe if you scrutinize special details in brighter areas you might find a teeny tiny bit more sharpness in the ArtixScan due to the fact that it has a few more pixels. In the deepest shadows, however, it loses completely.

As before, use the zoom-out function in your browser to study the images in less than 1:1 so as to judge the general perception of grains and deep shadow performance. Again, the camera scan wins! It looks cleaner with better blacks and less murky. If you look at the building in the middle with vertical stripes, I think the difference is startling! The camera scan is clearly superior in this respect!

In summary, I think it is pretty obvious that the camera scan wins! And this is with a camera that has a modest 18 Mpx sensor from 2009, which equates to 3657 ppi for a 24 x 36 mm slide!

More comparisons

As before, all images are clickable and open in popup-windows. At the bottom of each window, text says what it is and its resolution.

The two images below are a camera scan and an ArtixScan respectively of a Sensia 100 slide. (The photo is the sunset at Selfjorden on the northwest side of Moskenesøya, Lofoten, Norway, in July 2007. The lens is an EF 200/2.8L USM)

The top image is the camera scan. It matches the slide extremely well as viewed on the light table!

The bottom image is scanned by ArtixScan in SilverFast 6.5, no multisampling, at gamma 2.0 as raw as possible ('48 Bit HDR Color'). Then my best color management profile for this scanner was applied (based on 8x sampling of an IT8.7 reference slide from coloraid.de). Thereafter I have lifted shadows 2% in Photoshop, then converted it to Adobe RGB color space. Then imported to Lightroom where it has had various tweaks to make it match the camera scan as close as possible so as to make it possible to objectively compare them.

The Camera scan is simply better in all aspects! It was not possible to tweak the ArtixScan version (in Lightroom) to be as good as the camera scan. Especially, note the transition from the yellow-red sun rays into the shadowed mountain slope – the ArtixScan has some unnatural red shifts in various darker spots. Look at the left side of the image, in the very dark mountain slope: the camera scan is significantly more detailed.

The leftmost image is the camera scan. It matches the slide extremely well as viewed on the light table! Also Sensia 100, shot in 2001.

The rightmost image was scanned in 2004 by ArtixScan in SilverFast 6.2, at gamma 1.0 as raw as possible ('48 Bit HDR Color'), then assigned the IT8.7 profile created by inCamera Photoshop plugin (also at gamma 1.0). In Lightroom it has had various tweaks to make it match the camera scan as close as possible so as to make it possible to objectively compare them.

Again, the Camera scan is simply better! It would be difficult to tweak the ArtixScan version to become as good as the camera scan.

Glassless or glass frames?

If glass mounted slides look clean and ok, you might think that would be the best – well preserved slide and completely flat thereby eliminating focus problems due to curved film. It turns out that this is not the case, because a microscopic coating on the glass surface can cause scattered light that results in clearly visible artifacts that looks like dust.

The following two images are 1:1 crops of a Kodochrome II slide (shot with a Nikkor 35 mm lens in 1973) scanned in the camera scan setup, first in a glass frame and then glassless.

The leftmost image is from a camera scan where the slide was mounted in a double-glass frame (in 1973). Its outer glass surfaces were carefully cleaned before it was camera scanned.

Despite the slide looked perfectly ok as viewed on the light table with a loupe – not at all contaminated or speckled or dusty – there was a very thin coating on the inner glass surface (like some kind of "mist" or fog) that could only be seen in sharp sideways light that reflected onto the glass surface at an angle.

That coating is there from manufacturing to eliminate potential newton rings as the slide touches the glass, but this coating can (at least over time) cause sideways light to scatter: At bright-dark edges, scattered light from the bright area reaches the neighboring dark area where it is scattered again and lights up as speckles. The result is an effect that looks like dirty bright "halos" around areas of transitions from bright to dark. Note that these particles are not dark (as normal dust would appear), but rather, they are bright with a color that comes from nearby areas.

The rightmost image is from a camera scan where the slide had been detached from the glass frame and put in a glassless frame. It is just so much better!

Finally, I would like to point out how well the camera scan captures the Kodachrome color transitions and even the grains! It matches the slide as viewed on the light table extremely well in all aspects!

Summary and conclusions

A camera scan using a Canon EOS 7D (18 Mpx) with a macro lens is superior to a 4000 dpi meticulously color managed film scanner – a Microtek ArtixScan 4000tf. It is superior in noise, shadow performance, color detail and color transition renderings. In mid tone sharpness, they are about equal. The ArtixScan has a few more pixels (4000 ppi whereas the camera scan corresponds to 3657 ppi), but it is very hard to see if that translates into more image details (but it might translate into somewhat more natural-looking grains).

Negative film might perhaps be a nightmare with camera scanning – I simply have no idea! It all depends on what software is available to turn a negative RAW file into a positive one. Negative film does not have as much problem with very high contrast as slides do. A film scanner's most critical parameter is its ability to handle dark and high contrast slides. But with negative film, this may not be as important. Therefore, I assume that a film scanner is most at home with negative film, where it is very convenient to exploit the batch capabilities of most film scanners.

My list of findings and conclusions

• The camera scan is superior in noise, shadow performance, color detail and color transition renderings. Images from the ArtixScan could never be tweaked to match the slides (as viewed on a good light table) quite as well as the camera scans did.
• The camera scan clearly shows more details or structures way down in very deep shadows with better blacks, where the ArtixScan simply gets murky. This is the most startling difference in quality between the camera scan and the ArtixScan.
• The ArtixScan gives different deep shadow color casts on each scan. It is mostly seen with multisampling with radically different casts in deep shadows. The camera scan is always consistent.
• The only downside with the 7D camera scan is that it has a slight tendency to banding in the very darkest portions. (But I am pretty sure that the Mark II version of 7D would do better in this respect.)
• The sharpest scanning is the single scan with ArtixScan without any noise suppression applied in Lightroom, where even the grains are rendered quite well. However, it is overwhelmed by its own noise. With an adequate amount of careful noise reduction applied in Lightroom, the resolving power in mid tones is about the same as that of the camera scan with its adequate amount of noise reduction applied. (Both had some sharpening applied to compensate the loss of sharpness that noise reduction always causes.) Multisampling with ArtixScan always degrades sharpness, sometimes very much, so it is not worth the effort. Better to do noise reduction in software rather than fiddling with multisampling.
• The camera scan method rarely misses focus, which the ArtixScan does at times.
• The camera scan handles non-flat bulging slides better than most film scanners because I stop down to f/8, which provides an adequate depth of field for glassless mounted slides. The ArtixScan sometimes gives severely fuzzy corners since it focuses in the middle of the bulging film frame.
• The camera scan is always geometrically consistent, whereas the successive scan lines in ArtixScan are not fully equidistant due to limited mechanical precision, which causes a kind of distortion, as if there is some warping involved with straight lines becoming somewhat wavy.
• The sensitivity to dust seems to be about the same between camera scan and ArtixScan; I am specifically thinking of those microscopic particles that are mostly tied directly to the film emulsion and which can't be wiped away easily. When batch scanning with ArtixScan, however, it is significantly more sensitive to dust since, while a frame is being scanned, the other 3 frames sits horizontally waiting to be scanned for several minutes, and they can easily pick up dust.
• With the camera scan, it is very easy to check on the monitor at the back of the camera to make sure the exposure is correct and that it did not miss auto-focus. If not okey, just expose again. With a film scanner, you will not discover the problem until you have imported the file into your imaging program. Only then can you start a new scan.
• The camera scan file is only a 22-24 MB RAW file whereas the film scanner yields monster files of almost 130 MB for a scanner that does 4000 ppi 16 bits per channel.
• The camera scan is completely independent of computers, interfaces, operating systems, software upgrades, licences... With any film scanner, you are completely dependent on the current interface it provides and its version (such as FireWire 400, FireWire 800, USB 2, USB 3, USB-C, ...). With the next computer, or the next operating system, you can easily be in trouble! You will need new adapters, new software upgrades, and might need support to solve the new bugs that crept up after the software or hardware upgrade.
• A film scanner's primary merit, however, is that most of them can do batch scans, thereby streamlining a process of scanning thousands of slides (or negatives). A camera scan can't do that.

◊◊◊◊◊