A coffee break discussion prompted the following question:

How many pixels would a file contain if an 8x10 digital back were available (color or B/W)?

Seems there was a National Public Radio story that a consumer camera available in 2001 will have a 10 megapixel chip as standard equipment, rendering images equal to any film-based effort. I was arguing that in a 1/15th of a second, an 8x10 piece of film can collect huge amounts of "pixel" information, probably dwarfing 10 megapixels. Anybody know?

Thanks in advance - Bill

-- bill youmans (bill@greatgrips.com), October 03, 2000

Answers

Well, here's my rough answer: in round figure an 8x10 negative is about 200x250 mm. If we say that a good B&W film is capable of resolving well over 100lpm (again, just round figures), then you need at least 200 pixels per mm. LF lenses won't match that resolution, but you'll still need it to replicate the tonal gradation.

That gives 200x200x250x200 = 2 billion pixels. That is 200 times the size of that 10Mpixel chip, and that is almost certainly an underestimate!

Not only that, but I think I'm right in saying that each pixel uses several bytes for 24-bit colour rendition, and it's clear that for an uncompressed image you will need several GigaBytes of storage. I hate to think what the power supply would be like.

So, if my arithmetic is right, and you're planning that week-long photographic trip into the backwoods with current digital technology, you'd better hire half a dozen pack mules and a portable generator!

Huw Evans.

-- Huw Evans (hgjevans@hotmail.com), October 03, 2000.

"rendering images equal to any film-based effort. "

Yeah, but I'm sure they weren't considering LF...as most people don't.

-- sheldon hambrick (sheldon_hambrick@hotmail.com), October 03, 2000.

Huw,

You're right about color rendition - that requires a minimum of 3 pixels (R, G, B). 24 bit color likely requires more, but this is beyond my specific knowledge.

I think I also recall that a 35mm negative contains roughly 25-30 megapixels of data, but don't quote me on that. If right, though, digital is still a long ways away from a mano a mano quality comparison with film at any format. I wonder when the dramatic improvements we have seen in digital will run out of gas in the face of practical limitations on miniturization, and whether this will be before or after digital exceeds chemistry in data storage capacity. It's easy to say now that digital will ultimately win out, but I question whether it's really that simple (see the human brain).

Wow, I fell into an unanticipated philosophical twist. ;-)

-- Chris Werner (cbwerner@att.net), October 03, 2000.

we can also look at this from the other end. it requires about 300dpi to achieve truly visually-satisfying reproduction (print) quality (for example, the reproductions in a high-quality fine art book). 300 x 10" = 3000 pixels across, 300 x 8" = 2400 pixels high - that translates to 7.2 million pixels. at 24-bit resolution, that is about 173MB file size. at 600dpi (to allow for 16x20 enlargements for exhibition purposes), you would be looking at file sizes of about 690MB. this level of image quality is within the capability of existing technology (well, i admit you need a lot of RAM), and i would guess variations of it will be common in the near future. on the downside, if kodak wont even make a b/w film in 4x5 readyload anymore because the market is too small, who in the world do you think will manufacture an 8x10 digital back? (i wonder how long it will be before the LOC accepts any form of digital imaging - digital files, digital prints, and/or digitally printed "negatives" or some other form of hard-copy "original", as part of its archival collections...)

-- jnorman (jnorman@teleport.com), October 03, 2000.

I have done some experiments digitising film to see what you can get out of it and does appear that the limiting factor is lens technology.

You can see the grain in Provia F at about 6000 dpi but above 3-4000 dpi the lenses I have used have not provided any more information.

At 3000 dpi this produces a 720 MegaPixel image (2 GBytes (24 bit)).

If you are interested the results of scans at 1600, 3000, 6000 and 12,000 dpi they are at:- http://193.113.131.213/pg/dpi/?lf2

Viewing the images requires your browser to run a Java applet. I will have a none java system by the end of the month.

Regards

-- Nic Benton (nic@twob.co.uk), October 04, 2000.

Juggling numbers doesn't actually tell us very much about image quality.
If you consider the fact that every grain in a film can only be 'on' or 'off', i.e. developed on not, then film itself is very much a digital medium.
Now consider the nature of a pixel: Sure, it covers a much larger area than a film grain, but it's also capable of showing a whole gamut of brightness levels, from absolute black to pure white. Which is really closer to showing a truly analogue tone scale; film grain, or pixels?

The way that the human eye sees things must also be taken into consideration. The eye accepts regular patterns of dots as a continuous tone, much more readily than it accepts the random scattering you get with film grain. A coarse grained photographic print has many more grains per inch than a good 133 to 200 screen magazine or book illustration, but the book illustration appears to be smoother in tone than the print. The regular matrix of pixels gives a better impression of truly continuous tone than film grain.
IMHO comparing film directly with a digital image, on a purely numerical basis, is like comparing apples with bananas, but if you want some numbers, I think we ought to start with the human eye.
The eye can resolve at best about 8 lppm at normal reading distance. Even if you increase this to 10 lppm or 20 pixels/mm, this only works out to 82 megapixels to cover a 20" x 16" print. The final viewing size is much more important than the negative or film size.
Or, if you want the hypothetical equivalent of film: We'd need at least 255 film grains to show the same tonal range as a single pixel, and this involves a film area equivalent to at least a 16 micron square pixel. (The pixels in consumer digicams are about 5 microns square BTW). This is about 2.5 megapixels/square inch; you work out the numbers, they're far lower than the gigapixels that have been bandied about previously.

Anyway, wait until we've seen the results from the new full 35mm frame-sized CCD camera announced at Photokina. (At last the digital design boys have realised that size does matter.) It's only 6 megapixels, but this is well in excess of the 2.5 mp/sq. inch that I mentioned earlier.
I think quite a few eyebrows are going to be raised by it.
Leaf, LightPhase, and the rest should start worrying about their future.

-- Pete Andrews (p.l.andrews@bham.ac.uk), October 04, 2000.

Hmmmm....I've often estimated this myself, and usually come up with about the same number - 4000 dpi. The Kodak DCS-660, with 3Mx2M array, would need to have twice the linear resolution to confidently capture the detail in a 35mm negative (were the array 36mm x 24mm, which it is not).

The 100 l/mm number would require 200 pixels/mm minimum. (In signal processing, we call this 2x point the Nyquist sampling rate.) This already gives over 5000 dpi. In practical signal processing, we'd oversample somewhat. I could even argue that 4 pixels/line pair were required. (For those who can visualize it, imagine sampling a square wave right at the transitions - you'd get no variation.)

I usually start this mental exercise with about 50 l/mm to compare with 35mm film, so my "required" resolution is less than what I just showed.

Any way you slice it, you find that the VERY EXPENSIVE, top-of-the-line sensors today aren't close to capturing the information that film can.

You can follow the "300 dpi for print" rules and come up with smaller numbers. Of course, when you choose a resolution, you have to know how big your biggest enlargment or cropping will ever be. I'm a proponent of capturing all you can when you take the picture. It's kind of like cropping. If you take too much image (too much information) when you take the picture, you can crop. Likewise, if you take too many pixels, you can always downsample.

OK. I want to blather on more, but I'll stop now.

-- John H. Henderson (jhende03@harris.com), October 04, 2000.

Check this site for a comparison of various digital cameras to each other, and to 35mm Kodak E100 slide film. It seems that the 6MP imager used in the Kodak 660 and the Phase One med. format back (also using the 6MP imager) captures more detail than 35mm film. Very interesting! http://www.imagingspectrum.com/portraitcomparison.htm

-- Ron Shaw (shaw9@llnl.gov), October 04, 2000.

Don't really think you can come to that conclusion from these "tests." The type of digitizing for the film, compression for display etc. don't give a true method of comparing detail. Plus the digital pictures were "sharpened" with software giving a bias towards appearing to have more detail.

The only way you could really make a comparison is to use a lens resolution chart, enlarge the film onto paper to show the area with the most discernable line pairs per millimeter, and then perform the same test for the digital camera with output to a Lightjet or like printer with no software manipulations to the digital file. This isn't a real test as there were no controls applied to the process. It's only a sloppy presentation of some comparisons masquerading as information.

-- steve (s.swinehart@worldnet.att.net), October 04, 2000.

Sharpening wont create detail that isnt in the file. I didnt say sharper (but they are), but do indeed contain more detail. Like it or not, digital is coming of age. Digital solutions for LF are taking over studio work by the droves. Pros who make thier living with the images they create are finding that digital is making that easier. See how you feel about digital imaging in another 5 years.

-- Ron Shaw (shaw9@llnl.gov), October 04, 2000.

Everyone seems to like to say 600 dpi is the starting point. But realistically 600 dpi will not give you a "photo quality image" when printed. It takes at least 1200 dpi to print an acceptable "print". So you can double all the numbers you favor. But it doesn't matter to me in the least if digital is comparable to, or exceeds the "clarity" of film. Photography is a craft unto itself. I view it much the same as any other - the results are more often treasured for their artistic value and the appreciation of the artisanship that went into the final product. I think of the comparison as being similar to mass production. Why would anyone want hand-built, solid oak furniture made by a talented craftsman when they can buy a mass-produced, particle board/veneered piece at half the price? Which is most appreciated, and which will be treasured for decades or longer? Digital is to film what video tape is to film. Imagine video tape on a 70 foot theater screen.

-- Matt O. (mojo@moscow.com), October 05, 2000.

No need to imagine videotape on a 70 foot screen, you can go and see it, practically anywhere, because movie theatres are rapidly moving towards all digital distribution.

The use of crude black and white bar test patterns tells us absolutely nothing about the percieved quality of an imaging system except its ability to pass crude resolution tests.
Waving figures of 300, 600, and 1200dpi about is meaningless, unless you tell us the context. Is that on film, in a scanner, in a final print, or what? Is that a dye-sub or light-jet print, or a crappy fixed-dot-size ink-jet?

-- Pete Andrews (p.l.andrews@bham.ac.uk), October 05, 2000.

"Sharpening wont create detail that isnt in the file. I didnt say sharper (but they are), but do indeed contain more detail."

While that is true, the sharpened image will have the visual appearance of being sharper which gives the illusion of more detail.

The problem with the comparisons is that the test methodology used is flawed.

1. We do not know the amount of magnification of the images.

2. The film was digitized. This alone negates the entire comparison process. We do not know if number of pixels it was digitized at was appropriate for the amount of magnification. Further, we do not know if the film scan captured the maximum resolution of the film! Hence the need to use a fixed, know target such as a lens resolution chart instead of an aribitrary comparison of a portrait.

For example, while I might get away with 3000 ppi scan for an 8x10 from 35mm, if I want to go to a 48 inch wide print, I will have to have the film digitized at about 11,000 ppi. The same holds true for examining small areas of film to look for details after it has been digitized as this is, in effect, an enlargement,

3. The only way to compare the true resolution capabilities is through controlled tests of the entire imaging "system," and not through poorly thought out comparisons.

As to your comment, "See how you feel about digital imaging in another 5 years."

I feel fine about digital imaging. I use it daily. I've had papers published by the Society of Photo-optical Instrumentation Engineers on image processing of digitized infrared video. I also use digital imaging in my personal photographic work. I just don't to buy into a presentation that is really pseudo-photoscience.

I test video equipment for resolution, imaging, dynamic range etc. with image quality evaluated before and after digitizing. From my perspective, the tests were not setup and carried out correctly as the total imaging systems (digital and photographic) were not compared equally because of the test methodology used.

-- steve (s.swinehart@worldnet.att.net), October 05, 2000.

Steve, you apparently did not check out the url. They give all details of the images, and the scanning resolution (5700 DPI) on an Imacon Precision II Film Scanner. You can see the grain (or dye clouds) in the scanned film image. Some of the skin texture is obscured by the grain in the scanned film image, not in the Phase One image. This is also an old CCD (at least 5 years old). They also give the magnification (in equivalent print sizes, both 11x16 size, and 16x24 size).

-- Ron Shaw (shaw9@llnl.gov), October 05, 2000.

I've checked out their website, and I stand by my statement that their methodology is flawed.

-- steve (s.swinehart@worldnet.att.net), October 05, 2000.

These images were not printed. These are the image files, so printing has nothing to do with it. As far as scanning the film, if enough resolution was used to reveal the dye clouds, I dont see what a further increase in scanning resolution has to do with it, as the resolution used shows the limits of the film. Regardless of your opinion of the methology, I find the results interresting, and speaks a lot for the state of digital imaging, especially considering it is still in its infancy. Check out the images on the Hasselblad web site from thier new digital camera (at only 4MP, the images are superb). It will be interresting to see the images from Kodaks new 16MP imager due out early next year.

-- Ron Shaw (shaw9@llnl.gov), October 05, 2000.

Check out the website of Foveon, Inc. of Santa Clara, CA. It shows an 8 foot (96 inch) tall photograph captured with a 35 mm version of the 16 megapixel sensor chip. It will blow your mind. Hassleblad will be making MF camera with an enlarged version of this ch

-- David Caldwell (caldw@aol.com), October 07, 2000.

The web page is http://aolsearch.aol.com/dirsearch.adp?knf=1&query=large%20formAT%20HO ME%20PAGE%20

-- David Caldwell (caldw@aol.com), October 07, 2000.

David:

That URL didn't work for me. This one did: www.foveon.net

-- Greg Lawhon (glawhon@unicom.net), October 07, 2000.