What megapixel would a Digital camera need to be to match or exceed the resolution of a 35mm? Is this type of digital camera coming soon?

-- Frank Cerutti (mrfrank@mailexcite.com), August 07, 2000

Answers

25 million pixels to equal Kodachrome.

-- T. Muhlstein (tom.muhlstein@sympatico.ca), August 07, 2000.

I'm probably baiting the traditionalists here, but I'm going to suggest that your question may not be valid, and the answer may not be that simple.

I'm going to instead make the statement that: "the digitally extractable full resolution of film is unimportant -unless you're going render for output in a size large enough to require, or make use of, that resolution!" Or simply put, it doesn't require a cannon to swat a fly!

Most photofinishing and high resolution digital printing is considered highly acceptable at a mere 300 PPI(pixels per inch) of resolution in both axes. In some cases, the images are only printed at a bit over 200PPI with a dye sublimation printing engine. If a 3.3MP digital camera provides 2048 x 1536 pixels, then it's quite capable of producing a 6.8x5.1" 300PPI image!(Let's be fair and call that a 5x7" folks...) If we were willing to stretch things a bit further and print at 200PPI, we could produce a 10.24"x7.68" image. Near enough an 8x10".

The point I'm trying to make here is a somewhat subtle one. While digital cameras aren't yet capable of producing the amount of resolution you could scan from a 35mm negative, they are quite capable of producing sufficient resolution for many tasks, certainly a 5x7" and maybe an 8x10" or two.

Frankly, I think the figure of 25MP for a 35mm frame of film is perhaps gilding the lily just a bit? To my knowledge, the maximum PPI that can currently be extracted from a 35mm negative is limited to around 3000. As I recall 2820PPI, to be more precise, which yields about a 3886x2914 image, or about 11.3MP -let's be "sports" and say 12MP. Whether you can truly get more than that out of a print scanned on a drum scanner I don't wish to argue due to there being too many steps in the process of capturing, enlarging-printing, scanning, etc. each with it's own optical distortions. At such resolutions, seemingly minor optical distortions can produce noticeable defects, so the quality of the resolution obviously comes into question. So, the point I'm making is that at maximum resolution a 35mm is probably more nearly limited to around 12MP. Perhaps less, with consideration being given to film grain effects.

Perhaps, the real questions people should be asking themselves are: "How large an output medium will I use?" and "How many PPI must I image at for the expected minimum viewing distance?" The answers to those questions will ultimately yield the number of pixels you need to capture.

What say the lot of you?

-- Gerald M. Payne (gmp@surferz.net), August 07, 2000.

Let's just take the question at face value.
I have a 2700 ppi scanner, which isn't quite capable of getting all the detail off the sharpest of my 35mm slides, and certainly not from the best of my B&W negatives. It's not the fault of the scanners optical system, there just ain't enough pixels to do the job. Desktop CCD scanners are available with resolutions of 4000 ppi, and by all accounts they are up to the job. Now I know 4000ppi doesn't directly equate to 2000 lines/inch or 78 line pairs per miillimetre, but that figure isn't unreasonable to expect from the centre of a 35mm frame.
Even if we say that we need 1.5 times the pixel resolution in a scanner to capture the actual resolution on film, then we still end up with a film resolution equivalent of about 2700ppi. 27 megapixels. BUT, the pixels we get off a scanner are true pixels of one RGB triad per pixel, not the interpolated 3 or 4 pixels per colour that digital cameras give out.
I don't think 25 megapixels is at all outrageous. Add to that the fact that the larger image area enables truly wide-angle lenses to be used, that optical aberrations with long lenses are less problematic, and the control you have over depth of field, and you begin to see why current digital cameras still have a long way to go.

wake up you digital camera manufacturers, and start using physically larger CCD arrays!

-- Pete Andrews (p.l.andrews@bham.ac.uk), August 08, 2000.

Sorry! I was thinking megabytes, not megapixels. Divide by 3.
8 megapixels is the minumum needed to match 35mm.

-- Pete Andrews (p.l.andrews@bham.ac.uk), August 08, 2000.

Uhhh.. GOOD film has far more than 12MP of real extractable data. Using a Nikon LS 2000 and Polaroid SS 4000 I can tell you there is a visible gain to be had with the higher resolution. I'm get a little over 19 million pixels off the Sprintscan at 4000ppi. The scans have greater detail than the LS 2000 and there is image data over 2700 or 2800 dpi reguardless of what the general consensus is. The Imacon Flextight and Scitex scanners with 5700dpi exist for a reason and althought the Polaroid certainly isn't in that league, I've made scans on the SS4000 that are simply not possible on a Minolta 2800dpi unit. The real issue at hand are probably three things, your desired output size, camera optics, and film choice. The Nikon or Minolta are certainly enought horsepower for 8x10 to 11x14 work so depending on your goals, the breed of 2700-2800 dpi scanners may fit the bill. Second, its pointless to filmscan if shooting pictures on a crap camera. Obviously a decent camera with decent lenses is a prerequisite for even bothering to scan film. Cheap lenses with soft optics will only result in crap scans, garbage in, garbage out. As far as film, I personally MUCH prefer Fuji to any other film (except Ilford BW) for a variety of reasons becasue of their grain structure and exceptional enlargement capabilities. Cheap film, especially cheap negatives will never have much potential, again its a garbage in garbage out situation. Negatives, though they have a larger exposure lattitude, just can't hang with Provia 100F slide film. If you carefully bracket your slide film and combine multiple exposures in Photoshop, you can get images with a 8-10 stop exposure range all from two well exposed slides. Digital camera have a long way to go to catch up to 35mm, and medium format (especially 6x7 at 4.5 times the size of 35mm) blows away 35mm so this could take a while for digital to catch up.

-- Cris Daniels (danfla@gte.net), August 08, 2000.

Cris, would those "visible gains" be based on resolution target testing coupled with blind testing, or just looking at a larger scan and nodding and saying, "Yup, that looks better. Phew! Now I feel better about spending the extra bread for that new scanner..."? ;-) Relax, I'm just pulling your leg to see if you squeak.

If you suggest that accepted real world data and gross visual confirmation is enough, then I'd suggest that for several millions of consumers the digicam has already replaced 35mm... ;-) Especially if you consider that despite the quality of 35mm film in the hands of a professional the average person is used to seeing prints equivalent to 300PPI or less. Much less when squinting at 3.5"x5s that appear to have been printed by a preschooler with cognitive difficulties and a burgeoning Crayola "Autumns" fetish...

Personally, I'm pretty happy with my interpolated 3.3MP CCD based camera, because 5x7" to 8x10" seems plenty for me. But, YES, PETE!, I'd be FAR happier with an 8-12MP non-interpolated CCD about the size of a half playing card with electron wells deep enough to drop in a Volkswagon... :-) A nice set of 35mm or LARGER lenses to go with would be nice too. So I guess that gets us back up to 24 million sensor elements anyhow... :-)

-- Gerald M. Payne (gmp@surferz.net), August 09, 2000.

Chris. It's well known by repro professionals that in order to get all the useful detail off an original it needs to be scanned or copied by a system with one and a half times the resolution. (as a rule of thumb) That's the reasoning behind my figure of 2700ppi = 8 to 9 megapixels as the equivalent of 35mm, since going much higher in a scan than 4000ppi merely gives empty resolution. It would be nice if camera manufacturers made those real triad pixels though.

-- Pete Andrews (p.l.andrews@bham.ac.uk), August 09, 2000.

And Pete deftly plays the Nyquist card... :-) (again!)

It's a good point really, it comes back to the simple idea that if you have grid of say 200 x 200 data elements, and a grid of 200 x 200 scanning elements there's still no guarantee the two will align properly. Double the resolution in the scanning array(800 elements instead of 400) and your chances of "accurate" reproduction without aliasing are much higher. Nyquist basically said that 2x's was sufficient, any more than that and you're just wasting your time. If you look at it in terms of lineal resolution you need to increase the resolution of each dimension of the array by multiplying by the square root of two, about 1.414 or more often multiply by 1.5 as Pete stated. (simpler math...) Which means that even a 4000PPI array tops out at about 2820PPI.

Actually, Nyquist was talking about sampling frequencies, and stated that the sampling should occur at a minimum of twice the frequency of the highest frequency component in the signal. Funny how some concepts seem to radiate through various disciplines. Kind of makes me wonder just how long it'll be before some squint eyed geek(genius, that is) will discover a unified field theory and make it work...

The preceeding is a public information spot sponsored by the Useless Trivia Department. ;-)

-- Gerald M. Payne (gmp@surferz.net), August 09, 2000.

Ok, I'm wondering if Imacon and Scitex engineers are just mis- informed and building 5760dpi scanners to screw all the poor digital imaging suckers out of $15,000. Maybe someone ought to fill them in on all the "digital theories" floating around here. After all, if only 2700-2800 dpi is our "limit", I guess Nikon really isn't coming out with a higher resolution film scanner this fall after all. I own a SS4000 and work with a LS2000, I know that when I scan images off the Polaroid and they are 58Mb compared to the 28MB Nikon, they contain greater detail when magnified and using the multisampling capabilites of Silverfast 5.1, very clean shadows for the price range of the scanner. I know I can print 13x19 @300ppi without resampling and that is not possible with the Nikon or Minolta. I know that when I calibrate the scanners with IT8 targets, I get exceptional color reproduction and predictable results. I own a mid-range Nikon SLR and get results superior to the Kodak DCS Pro digicams, I've used them and have prints off a 8660 Dye sub. 9 megapixel cameras would be nice but again, medium format contains MUCH more data than that. If you want to read about a talented person using current digital imaging hardware, check out www.luminous-landscape.com and ask Michael why he bought an Imacon and owns a Polaroid SS4000, why he can see apparent differences in resolution you claim are scientific nonsense. I currently use or have used the equipment I'm talking about so its not like I'm speaking from a "booksmart" idiot perspective.I have color correct lighting in my studio, neutral paint, monitor hoods for my computers, air purifiers to keep down dust, a custom viewing booth for prints, and MacBeth targets to shoot to watch my color. My findings are real and actual, for me they produce the desired result and people I show my work to often ask what equipment I use, so I help them to the best of my knowledge. If you want to believe there is no data beyond 2700ppi, whatever...

-- Cris Daniels (danfla@gte.net), August 09, 2000.

Cris, you're still not getting it are you? I'll type it more slowly.
To get 2700 dpi worth of information from a 35mm original, you need to scan at 1.5 to 2 times the resolution.
Therefore, if all the information in a 35 mm frame can be retrieved by a 4000 ppi scan, the original only has 2700 ppi worth of information in it to begin with. Can't make it much plainer than that.

-- Pete Andrews (p.l.andrews@bham.ac.uk), August 10, 2000.

I wonder how many prints are focused finely enough to justify the scanning resolutions being talked about

-- Dan Morris (dmorris@aol.com), August 10, 2000.

Hope no one minds a stranger jumping in here. I recently bought a 3megp DC based entirely on the final customer. The publisher I'm targeting requires a minimum of 300X300 jpgs and readily accepts digital work shot at that resolution. Sure you can extract more info (ok, tons more) from a slide but the publisher can't print it. I would have to say that right now, in this area, digital is competative with film.

-- Richard Kellaway (pcservices@softcom.ca), August 11, 2000.

Using a Nikon 990, FINE setting, Photoshop 5.5, resetting image size to reflect 300ppi, Genuine Fractals and/or Photoshop's resampling engines, an Epson 2000P printer I have produced an image a professional newspaper photographer had trouble believing wasn't a "silver" print.

the defense rests.

FHerzog

-- Frank Herzog (Fherzog@mindspring.com), August 13, 2000.