The Pixel Race

I’ve long thought that the resolution of most digital cameras has reached the point at which it’s overkill, and there are a lot of other improvements that the camera needs. Unfortunately, the marketing people at Canon don’t agree:

Canon engineers are being held back from developing new sensor technology by marketing departments in a “race for megapixels”, claims an employee of the Japanese photography company.

The employee told Tech Digest that Canon have the technology to “blow the competition away” in terms of image sensors, but are instead being asked to focus on headline figures like the number of megapixels a camera has. When asked for his opinion on the Canon EOS 5D Mark II, which we covered this morning, the employee said:

“I am hugely disappointed because once again Canon engineers are dictated by their marketing department and had to keep up with the megapixel race. They have the technology to blow the competition away by adapting the new 50D sensor tech in a full frame format and just easing off a little on the megapixels. Although no formal testing has been done on the new model yet, judging by the spec and technology used, it just seems to be as good or as bad as the competition – not beating them by a mile (which we used to).”

I’d rather have more speed and better S/N ratio myself.

There’s an amusing discussion of this, and the perennial war between marketing and engineering, including examples from Dilbert, over at Free Republic.

24 thoughts on “The Pixel Race”

  1. The industry knows this is a problem. But one problem that’s unsolvable is that consumers don’t like to learn too much about their products. They want simple metrics – megahertz, MPG, 5-star safety, etc.

    That’s why the Cameraphone Image Quality (CPIQ) Association is trying to establish repeatable, objective and robust tests of image quality. They’re in “Phase 2” (whatever that means) of establishing and propagating those standards. Once those have been established we’ll have a better simple metric to market cameras with.

    http://www.i3a.org/technologies/mobile-imaging/cpiq/

  2. seems like the same things going on with Sony. Last year I bought a pocket Camera for my Wife for a press. Did the research and got a very highly rated pocket cam from Sony. A couple months ago I bought a cam for myself. In the past year, Sony replaced that top rated 5 MP, with a identical looking 6 or 7mp with so so rating given the worse image quality. And replaced that with a still lower rated 8mp in the same form factor. Super high res – but with static adn opur colors.

    I wound up buying the old model no Amazon, at a significantly high price then they were when new – or the new 8mp version is now!

    I see marketing problem?

  3. Depends on whether you’re going to make prints or not. Last I checked, which was within the past couple of months, no monitor available for less than $1000 will display more than ~2.3 MP, so almost any old digital camera will do. But an 11″x14″ print at 300 dpi is ~14 MP.

  4. What Jay said. It’s a matter of what you want to use the pictures for.

    Film is still better for professionals and that’s the ultimate target market they’re pushing for.

    My Phone has better MP than my first digital camera but the images are pretty poor because the software and the optics aren’t up to is. OTOH, I’ve seen the output from 2.1MP Sony Ericsson SureShot phones and they’re excellent.

    It’ll probably settle down eventually.

  5. -On photo discussion boards that I read, people have been complaining for years about new camera models that have more megapixels and lower image quality than the models they replace.

    -The more megapixels, the bigger the file size. This is a real cost for users.

    -Other things being equal, the more megapixels the more noise. This may be obvious to Rand but apparently not to most camera buyers. What’s odd is that the camera companies (with the possible exception of Nikon on their D3 and D700 models) use the more-megapixels gambit on their high-end cameras, which are marketed to people who should know better. I would much rather have fewer MP and less noise.

    -Other industries are worse. Expensive bicycles seem to be designed almost entirely by marketing people. Consequently a lot of the new technology is inferior (less durable or solves the wrong problems) to the stuff it replaces. But the customers eat it up so who cares.

  6. The resolution of a Digital SLR will never match that of a film camera. Film resolution is effectively the size of each photosensitive molecule, which is to say, much smaller than any single pixel on any image sensor ever made or probably ever will make. Of course, the question is, who really needs tech at that level anyhow? Perhaps someone.

  7. My old 6 mp Canon EOS 10D makes lovely prints up to 13×19″ if the original image is shot cleanly with plenty of light. It is very very noisy in low light, though.

    I’d love to see Canon’s latest tech in, say, a 6 mp chip that can shoot at 6400 iso but has noise levels of my 5 yr old 10D at 200 iso.

    They can do this. But they won’t.

    Meanwhile, I keep shooting because I don’t NEED 12 megapixels. Or 20. This camera is GREAT. The large prints look GREAT.

  8. You’ll laugh, but I have an ancient Canon PowerShot A10 that is a measly 1.3 MP, and I love the thing. It uses regular AA batteries, and I can email the full resolution pictures. The pics are much, MUCH better than the ones my iPhone takes, and it’s 2 MP. Seriously, there was a request by a pro photographer for interesting pics for her blog, and she took one I shot with the A10 and used it! It was a pic of Thomas Jefferson’s brick outhouse at Poplar Forest. LOL!

    My only worry is that it will wear out, because there’s nothing comparable to replace it with now (That I’m aware of). I think I’ve had it for, like, seven years or so.

  9. I upgraded from a D70 to a D300 and I find I get less noise in my pictures even though the D300 is a 12MP vs. 6MP for the D70. I read reviews carefully to verify the better noise handling before buying. In side by side comparisons, the D300 has clearly better image quality all around.

    On the other hand, I used to have Sony T-1, but it was lost while on loan to a friend and replaced with a T-10 and it seems to have noticeably more noise under anything but optimal lighting.

    I very definitely like the increase to 12MP from the D300. While it’s overkill in theory, in practice I take a lot of action shots and the the extra pixels give me a lot more freedom to crop, which makes getting the framing right much easier. One can think of it as providing a 40% digital zoom with no loss of quality, which is nice. I can see, though, that for casual consumers who don’t bother with post processing, it’s probably not worth it.

  10. “Jay Manifold wrote: But an 11″x14″ print at 300 dpi is ~14 MP.”

    Not it’s not. PDI and PPI are two totally different things. Printers drivers do not map pixels to dots on a 1:1 basis. Not even anywhere close to that.

  11. Panasonic’s new Lumix LX3 (and it’s Leica DLUX-4 twin) backs off of increasing MP count in favor of better noise control and image quality. It’s not much, but at least someone in the industry is considering it.

    The resolution of a Digital SLR will never match that of a film camera. Film resolution is effectively the size of each photosensitive molecule, which is to say, much smaller than any single pixel on any image sensor ever made or probably ever will make. Of course, the question is, who really needs tech at that level anyhow? Perhaps someone.
    Film resolution isn’t at the molecular level, it’s at the level of the silver halide crystals with which film base is coated. Those crystals vary in size depending on film speed and composition, but a) digital resolution parity with film isn’t just possible, it’s around the corner, and b) if you take anything more than snapshots you need the resolution to enlarge or crop.

  12. I forget where I read it originally but I came across one story of how a photography professor at some University set up a booth and displayed the same picture at different megapixel resolution but didn’t tell anybody which one was which.. People would submit guesses as to which photo was which resolution. The lowest was 1 megapixel on up through 10 megapixel. From this informal study it seems that the human eye really looses it ability to determine resolutions at anything beyond 5 megapixel. People were routinely saying the 10 megapixel image was the anything but. He only had one person that guessed all of them correctly but it was someone working on their doctorate in some visual imaging field.

  13. This isn’t new – it’s what killed Foveon.

    They’re the folks who used a single sensor for RGB instead of separate sensors for each color. (Foveon’s sensors are stacked vertically.) BTW – The “pixel count” is the sum of the number of R, G, and B sensors. They use interpolation to get the other colors at each sensor.

    That single sensor is larger than a comparable array of single color sensors (1 R, 1B and 2G IIRC). However, it also has better accuracy and low-light characteristics than the separate sensors and all of the data is sensed; nothing comes from interpolation.

  14. Jay Evans wrote: “PDI and PPI are two totally different things. Printers drivers do not map pixels to dots on a 1:1 basis. Not even anywhere close to that.”

    OK, what’s the ratio and what kind of pixel count would be needed for a crisp 11″x14″ print? Is there variance between printers? Are they catching up with camera pixel counts? Should they?

  15. A few years ago we intentionally went with the Fujifilm V10 (5MP) over similarly priced higher MP models simply because of the excellent SNR offered by the finepix chip. The ability to capture low noise usable impages at ISO 400+ and natural lighting is often under appreciated.

  16. Jay: I don’t know the ratio, but Ken Rockwell (who would know) suggests that a 3MP camera will produce 12×18 prints that look just fine to the naked eye, and professional art print sellers can manage 13×19 prints from a 6mp D70.

    I am absolutely confident that with a crisp input image and a good printer (ie, the hardware not getting in the way), one can get a crisp 11×14 print from any consumer-level DSLR and any P+S that can produce a decent source image (ie, older ones that aren’t so noise-prone, like my A520).

    (I never print mine, but my coworker has a D40 and has produced excellent 8×10 prints with 6MP.

    I think they have just slightly less effective resolution that I get from a 6×6 shot on Portra 160 – and that’s not exactly comparable to a 35mm camera.

    And God knows the cost per shot is almost zero, in direct contrast to shooting 120 roll film. There’s a reason I never use my MF cameras lately…)

  17. Pixel count doesn’t matter much, but sensor size does. In anything but bright sunlight with slow-moving things the larger sensors they put into DSLRs just blow away the tiny ones in pocket cameras.

    In the mid 90’s I was doing a lot of computer consulting work (Postscript) for a top pre-press house. They got a lot of work in the form of 35mm negs, which they would scan on a $200k drum scanner. It was capable of silly high resolutions, but they told me they never scanned 35mm film at more than 3000×2000 pixels because that was all the information the film really held.

    It was at that point that I decided I was going to replace my old Olympus SLR with a digital version once a 6 MP DSLR got down to a reasonable price point. That happened with the Nikon D70. I was a bit slow off the mark and got a D70s towards the end of 2005.

    I see absolutely no reason to upgrade to more megapixels. And more dollars I have to spend on cameras in the near future will go into lenses, not more pixels.

  18. Funny you bring this up Rand. We have the exact same problem with focal planes used in missile warning sats (and we aren’t even trying to grab market share). People keep demanding more funding for expensive focal plane development to get more megapixels, when more pixels will hardly provide any improvement in meeting missile warning needs. The same money could be much more effectively spent in sensor architectures and supporting hardware R&D – but we have continued to blow money chasing the next biggest toy instead of producing a kick-ass system with the technology we have today. I thought the similarity was amusing.

    Fortunately things are coming around.

  19. Using 2Mpixel 1/2″ Firewire cameras I found that the pixel spacing was below the Rayleigh criterion (search wikipedia) for a perfect lens (~9mm f/4 Cmount). It gets worse at higher f-stops. Has anyone compared the angular resolution of their lens to the pixel spacing on the new high end cameras? Its pretty simple geometry. It seems that most compact cameras have long since passed this.

  20. You can look up the pixel size or estimate it with diag imager size in mm * 0.8 / #of pixels in X.
    The smallest point a lens can make on the focal plane is f-stop * 5.62e-6mm.

  21. I don’t see why people are optimistic that this stuff will “settle down” or work itself out. Look at PCs and printers and just about everything else in the electronics department. They’re all festooned with stupid stickers, they all come loaded with crapware and it’s been a mess for over twenty years.

  22. >they never scanned 35mm film at more than 3000×2000 pixels because that was all the information the film really held.

    I concur. I sometimes used Kodak Photo CDs in the 90’s, and at 2048×3072 pixels you could notice the grain in a 35 mm ISO 100 negative.

Comments are closed.