Pixels Vs Film
A Draft Article by Brad Templeton
How many pixels are there in a frame of 35mm film? A print?
This is a somewhat controversial question, and there are many possible answers. Film is an analogue medium, so it doesn't have "pixels" per se, though film scanners have pixels and a specific resolution.
Today the one thing most people agree on is that it's a lot more than any current consumer digital camera. The debate is about how much resolution the digitals will have to reach to start matching the film.
The very short answer is that there are around 14 million "quality" pixels in a high-quality 35mm shot. That's a shot with a tripod, with a decent lens and a quality, fine-grained film, in decent light. There may be as few as 4 million "quality" pixels in a handheld shot with a point-and-shoot camera or camera with a poor lens. And of course if focus is poor, or light is poor, or the camera was not held steady, the number will drop down below the 1-2 million pixels of the modern consumer camera. Of course, one can have a bad shot with a digital camera too, not using all its resolving ability.
It's important to note another key difference. Film, as an analogue medium, does not record just 256 greyscales or the corresponding 16 million colours. And film scanners, even doing just 8 bits per colour, get 24 bits of data for every single pixel. Today's digital cameras only get 8 bits of data for each pixel and they guess (interpolate) the other 16. So the colour accuracy for even a scanned film image is much better than the modern digital camera. Good film scanners can also extract more than just levels from 0 to 255. They can often go to 12 bits (0 to 2048) to detect much more detail in shadows, and provide more contrast. As such a film scanner gets as much as 36 bits of information for each pixel, instead of 8.
The film itself tends to be able to hold around 1000 to 1 contrast range. Quality slide film can handle more levels, though over a slightly narrower range. Generally one desires at least 12 bits per colour to represent it. Your eye, by widening and closing the iris, can sample an astounding (eye-popping!) 7 decimal orders of magnitude of range of contrast, which would need at least 24 bits.
So there is a lot of information in film. However, not all of it is usable information, which causes the debate about the equivalence in pixels. Film is made up of chemical grains. The more you blow up film, the more you start seeing noise caused by those grains, and eventually the very clumping of the grains themselves. Of course some are bothered by the grain more than others.
The bank of the Colorado River near Moab, Utah. The late-afternoon sun reflected off the opposite red-rock canyon wall creates other-worldly colours.
Photographed with a Hasselblad XPan and 90mm lens on Fuji Velvia, this 100MB+ file (scanned at 3,200 PPI with an Imacon FlexTight Photo scanner) is filled with exquisite detail. It looks like nothing in a small low-resolution reproduction like this, but in an 18" wide print it sings with texture and pattern.
This is an example of resolution and detail exceeding the medium (a small web display) and thus failing both as art and as record, since all of the subject matter that makes this images work ‹the tress and rocks ‹ actually fall below the resolution limits of this display medium.
The finest films (which are slow and best with sunlight or flash) have very fine grain, and in many cases, the limits of the lenses blur the
image before the grains start causing too much trouble. However with a decent lens you don't have to blow
up too much before you see the grain. Of course, digital images, once they get to high resolutions, will also be
subject to lens limits (and they already are on cheaper cameras.) They are actually more subject because CCDs are currently much smaller than 35mm.
These films, with good lenses, are capable of resolving as much as 7000 pixels (3500 "line pairs") over the width of a 35mm frame -- about 5000 dots per inch. However, at that point, while they can resolve "line pairs," the image is pretty noisy. The lines are not resolved as straight, sharp-edged entities, but you can tell there is a white line next to a black line.
There is more information to be extracted even at this fine resolution, but the deeper you go, the more noise you also extract.
To make the image not look "grainy", you need to pull back. Subjective tests suggest this is to about 3200 DPI, or around 4300 pixels. For a 4:3 frame, that means a bit under 14 million pixels. (Of course some people don't mind grain as much as others, so your mileage may vary. Also, if you can get a scan that good, digital techniques can reduce the visibility of grain and extend the resolution of film.)
What this means is that a 4000 x 3000 digital camera would produce a shot as good as a quality 35mm camera in most conditions -- provided you could get more than 8 bits per pixel. You could blow up the 35mm shot a little bit more and see a little bit more, but only at the cost of producing a grainy image.
Prints also are analogue output. In theory a print can have all the information of film, however photographic paper tends to only be able to hold a range of 100 to 1 in contrast. That's less than a monitor can. In addition, the printing process is not perfect, and often blurs an image. Typical lab prints don't seem to store much more than 200-250 pixels for each inch. Quality labs can do better. Thus a 5x7 print probably is similar to a 1400 x 1000 digital image (if the digital image has enough bits per pixel.) It's rarely more than the 2100 x 1500 that 300 dpi would imply.
It's always better to scan from print film or slide film. You get far more contrast, and far more detail, and it's not second generation.
Makers of digital printers play lots of games with their resolution. The "dots" they speak of (when they talk about a 720 dpi resolution) are dots of single colour ink. A pixel, on the other hand, is a dot capable of the full colour range. You need lots of dots to make a specific colour and not look spotty. To render a pixel well can require scores of ink dots. In the end, the goal is a "continuous tone" image at a given number of pixels per inch. Most printers can only simulate continuous or near-continuous tone. (Digital film recorders, or photographic paper recorders, can do near-continuous tone.)
This article is © 1999 by Brad Templeton.
Photographic illustration by Michael Reichmann, © 2000