Digital Photographic Print Processes
With the ever-increasing popularity of digital cameras, I’ve been thinking about the best way to make hard copies of digitally-taken photographs. It seems that there’s a match between digital cameras and continuous-tone output devices, such as dye-sublimation printers. In fact, I suspect that the quality of printed image that can be achieved this way should be able to surpass that of traditional film images printed electronically. (I won’t comment on the relative quality of traditional film images printed traditionally.)
Pixels Versus Crystals
Consider the following.
A continuous-tone printer is able to produce millions of different shades of color on a single spot on the paper. That’s what makes them so cool. Dye-subs are an example of this. So, you don’t need incredibly high pixels-per-inch resolution to be able to render a detailed image. Color depth is important.
On the other hand, on-or-off printers, like ink jets or laser printers, can only print or not print a color dot on a single spot on the paper. In order to render shades, they must dither, which reduces the effective resolution significantly. That is, if you’re using dots that are 1/1000 of an inch across, your effective resolution for continuous-tone images might be more on the order of 125 dpi, since it would take 8 dot spaces across in three colors of ink to render 24 bits worth of shade.
OK. All that’s in the image production end. Now, consider image capture.
Traditional cameras using chemical film use a technique where photons chemically convert crystals of photo-sensitive material so that they turn dark. Any given crystal is either dark or removed (during the development process). The resolution of the film is determined by the size and density (and to some extent, shape) of these crystals. The same is true of color images; you just have three overlapping layers of crystals which are linked to dye molecules. This should remind you of ink jet printers. It’s really quite similar. The primary difference, neglecting scale (silver halide crystals are smaller than ink jet dots) is the arrangement of the spots of color or lack of spots: An ink jet’s dots are arranged in a nice, orderly, rectangular dot matrix, while the silver halide crystals on film are more randomly shaped and arranged. The latter property tends to hide the grain of the crystals a little. You don’t end up with Moire patterns, for instance.
On the other hand, digital cameras use CCDs to measure a quantity of each of the component colors at each pixel. That is, a single spot on the digital "film" can record a range of shades. These spots or pixels are arranged in a rectangular matrix, each pixel having complete (deep) color information. This would seem to be a perfect fit for continuous tone printers, which deal with images exactly the same way.
Comparing by the Numbers
Let’s start with a reasonable-quality digital camera, such as the Nikon D1. Ignoring the pixel color-offset problem, and assuming you don’t use JPEG or other lossy compression to store the images, then you have a 3008 x 1960 pixel image. There will be no artifacts from the optical system of a film scanner, no film grain, and no dust.
So, that’s basically equivalent to a 2000-dpi scan from 35mm film, except that it’ll be sharper. That doesn’t seem radically different from the 2700 dpi scans I can get from my Nikon LS-2000 film scanner.
Printing an image of this size at 20- by 30-inches would come out to just about exactly 100 dpi, even if the native resolution of the output device were 300 dpi. Realistically, I suspect that a 100 dpi continuous-tone printout at that size would look pretty darned good. Consider that you wouldn’t be looking at it from a foot or two away, like you would a 4x6—you’ll be a few feet away at the least. I doubt your eyes have greater resolution than that.
Now, I’m talking about fairly high-end digital imaging products, here, such as Nikon D1. But, that said, I don’t think I’d consider digital cameras viable until they had around that resolution. It’s for this reason that I haven’t seriously looked at buying a digital camera, yet. With even today’s point-and-shoot film cameras, the resolution of the image you get back is sufficient for blowing up or scanning in or what-have-you. Until digital cameras don’t compromise image quality, I’m not terribly interested. :)
Other Considerations
Another interesting thing is the computer display in the middle of the digital imaging process. When you display an image on your computer screen, you’re seeing a square grid of pixels, each of which has color depth. This is perfect for images from digital cameras, and for output to continuous-tone printers, but it doesn’t match film or ink jet printers very well. Now, of course, that’s not a problem for me. I scan images from film using a film scanner, which converts from colored spots to continuous-tone pixels, and loses resolution along the way. (Consider it this way. If you scan at a high enough resolution, you can see the grain, which limits your resolution. If you try to remove the visibility of the grain, you’re necessarily going to lose picture information, because you’re trying to convert from one resolution domain into another. If you don’t scan at a high enough resolution to see the grain, then you’re below the resolution of the film to start with.) But then, I have the image in continuous-tone pixels, and my plan is to print them out on a dye-sub printer, so I only lose once. Scanning from film into a computer and then printing out on an ink jet seems like the worst combination.
This also makes you think about the necessary dot density of ink jet printers to get decent results with images taken with digital cameras. The computer part of the process doesn’t lose you anything here. But how many dots do you need per inch in your ink jet output to avoid losing any of the resolution of the source camera?
One more interesting issue is the offset of red, green, and blue in images. When you scan an image from film, or when you print it on a continuous-tone printer (or an ink jet, for that matter), the color components of the image are exactly aligned on top of each other. (OK; for output devices, it may not be exact, but any error is going to be in a random direction, and vary from printout to printout, assuming the device is properly adjusted.) Monitors and digital cameras that take an image in one shot using CCDs, on the other hand, necessarily have their red, green, and blue components offset by a small amount. That’s because on a monitor, the color components of the pixels are not exactly on top of each other, and the light-receptive cells of the CCD camera for each color are similarly next to each other. The effect is that the red component of the image is sampled from a place just to the left (say) of the green image, offset by about one third of a pixel. When we print this out on a dye-sub printer, the color components are being shifted slightly with respect to each other.
Of course, since this shift is about a third of a pixel, the amount of the shift is decreased linearly with increasing resolution. And at the resolutions we’re likely to be doing digital imaging, we’re unlikely to notice.
However, it’s worth pointing out that this effect does decrease effective resolution, and the ability to eliminate this problem is one of the selling points of three-pass digital cameras.
- Geoff
11 Jan 2002