Digital Camera Resolution versus Film Scanner Resolution
Digital cameras create digital images composed of lots of tiny pixels. So do film scanners by digitizing your existing slides or negatives. A good digital camera will create images between 6 and 12 megapixels while a good scanner can easily produce files with more than 20 megapixels. So does this difference have any bearing on which is better?
The answer may not be quite what it seems at first glance.
Reality isn't made up of discreet little bits like pixels. It tends to have continuous detail throughout, with clearly delineated edges where they are supposed to be, and smooth transitions of tone where needed. Life doesn't line itself up on an orderly grid of rows and columns, so in order capture it on a camera's CCD or CMOS sensor, we have to impose one on it. Imagine looking at the world through a typical window screen. What you would see through each hole in the screen could be thought of as being equivalent to one pixel in your camera's view of the world. But each photosite on your sensor has to average what it sees and record just a single value. It can't discriminate any detail smaller than the size of each pixel.
A film camera records images in a fashion that is surprisingly similar to a digital camera given the immediately obvious differences inherent in the two formats. Rather than regular rows and columns of pixels though, film has grain. A typical film is comprised of three dye layers, one each to record red, green and blue light. Each of these layers is completely covered in film grains made from silver halide crystals in a random arrangement. Resolution on film is limited to the size of a grain much as it is to the size of a pixel with digital.
Comparing the size of photosites in digital cameras to that of film grain yields some interesting results. My Nikon D2x has a sensor that measures 23.7 x 15.7 mm (the standard Nikon DX format) and is capable of creating images that are 4288 x 2848 pixels (12 megapixels total). Dividing this out yields a pixel pitch of around 5.5 microns (a micron is one thousandth of a millimeter). My old Nikon D100 had only six megapixels instead of twelve and recorded images with a maximum size of only 3008 x 2000 pixels, also on a DX-sized sensor, which works out to about an 8 micron pitch. Of course, a digital camera records only red, green or blue at each photosite in a Bayer mosaic pattern and interpolates the remaining values to the true resolution possible is dependent on the quality of the interpolation algorithm used. Professional slide film has a grain size of between 8 and 11 microns (Fuji Velvia 100F has an RMS grain size of 9). Most photographers who have shot both film and digital end up being surprised how similar the resolution between the two ends up being. Current six megapixel and up DSLR bodies are capable of results that are roughly similar to that of quality slide film.
But what about slide scanners? As its name implies, my Nikon Super Coolscan 4000 is capable of scanning at resolutions up to 4000 dpi (technically spi, or "samples per inch," but everybody calls it dpi). A 24 x 36 mm frame of film scanned at 4000 dpi will yield an image of over 20 megapixels. At first glance, this should mean that scanned film images are better than digital camera images. After all, they have considerably more megapixels. Even a 2700 dpi scanner will give you a nine megapixel file.
Scanners have to have such high resolution though in order capture all the detail the slide has in it. Photographing a scene on either film or digital involves some loss of detail in the translation to photosites or film grain. Any pixel or grain that "sees" a solid color can accurately record what it needs to, but any that need to record half of one thing and half of something else are out of luck. Things have to be averaged out to get something that can be recorded. Edges get softened to a degree and adjacent colors merge since the size of the photosite or film grain determines the smallest detail that can be recorded. If you then scan a frame of film, you subject it to this sort of process a second time, further degrading the image unless you scan at a resolution high enough to resolve the actual grain structure.
Indeed, something called Nyquist's Theorem states that you would need to scan at at least twice the resolution to get the information in the original. Since Velvia 100F has an RMS grain of 9, we'd need something that can resolve at least half that, or 4.5 microns to get the information the slide has to offer. That works out to about 5600 dpi and scanning at this would yield a file with over 40 megapixels. And it gets worse than that potentially for scanning film since grain is both irregularly sized and irregularly spaced. The RMS grain figure quoted is an average so the scanning frequency may need to be even higher than double in some cases. Ouch.
As mentioned, an image from a digital camera really only records one of the three color channels at each photosite and interpolates the rest. But as can be seen here, scanning slide film presents an even bigger hurdle to overcome in order to make every pixel count. Clearly file size by itself doesn't tell you everything about resolution. If pixel count was all that mattered one could get more resolution by simply upsampling in Photoshop. And we know that won't work. True resolution can't be greater than the pixel count implies, but it can easily be less. And often is.
Shown below is a series of different versions of the same photo. The first one shows a close-up portion of an image I shot of Hallett Peak from Nymph Lake in Rocky Mountain National Park in the fall of 2004.
The second version shows how a simulated digital camera might see this scene. The image has been reduced to regular rows and columns of pixels.
The third version was created using the Crystallize filter in Photoshop in order to simulate how film grain might see Hallet Peak. The detail that remains is different, but roughly equivalent to that in the version above it.
The last version shows the results of applying both processes and represents what you would get by scanning film at the same resolution that our simulated camera shoots at. Even more loss of resolution is evident here than either process caused by itself. Indeed, the only way to avoid further degradation would be to scan it at a resolution approaching what the images here are actually presented at. That is to say, since what I have here are 300 x 300 pixel images of each version, we could avoid further degradation to the simulated film image by scanning it at 300 x 300 pixels. Short of that, we risk further loss of perceived resolution as discussed above.
Portion of original image from Rocky Mountain National Park
Simulated digital camera version
Simulated film version
Simulated scanned film version
So which is better, film or digital? Well, of course, whichever one gives you the results you are after is. Better for you and your needs at least. But if you've been waiting for a camera that can give you an image as big as you can get from your film scanner, it may be time to rethink where you set the goal posts.