Slashdot Mirror
Beyond Megapixels
TheTechLounge points to this "first of a three-part series of editorial articles examining current digital photography hardware, as well as the author's views of what is to come." It boils down to the excellent point that pixel count alone is not the way to evaluate digital camera capabilities.
The biggest determining factor to me in buying a good digital camera is the optical zoom. With so much focus put on the number of megapixels and digital zoom (which, in my opinion, is better done in Photoshop anyways), the optical zoom is too often forgotten and hard to find in most "affordable" digital cameras. Without the optical zoom, one is limited to the same twelve-foot-away pictures that is great for people who only want to take pictures of friends and family standing in front of things, but is really useless if you want to get a good close up.
For example, this picture I took with my decent megapixel digital camera, my first time using it was a terrible disappointment because it was a great shot ruined just based on my not having the proper optical zoom capabilities.
(And my mistake in buying a camera that I thought would be top of the line, and stupidly didn't notice the difference between digital and optical zoom, this being my first move off of traditional cameras.)
*-*-*-*-*-*-*-*
"We are Linux. Resistance is measured in Ohms."
Check out this link. It details a bit on how the spirit rover only has a 1 megapixel camera on board, yet delivers IMAX quality images.
From the article: "NASA's Spirit Rover is providing a lesson to aspiring digital photographers: Spend your money on the lens, not the pixels. Anyone who has ever agonized over whether to buy a 3-megapixel or 4-megapixel digital camera might be surprised to learn that Spirit's stunningly detailed images of Mars are made with a 1-megapixel model, a palm-sized 9-ounce marvel that would be coveted in any geek's shirt pocket. Spirit's images are IMAX quality, mission managers say. "
The camera sitting at the extreme of the low megapixel, high quality spectrum is the Sigma SD10, which is the only camera to use Foveon's x3 sensors to capture three colors per pixel. This results in a very high quality image, even though the total pixel output is ~3.4 megapixels. I would like to see some of the other major players put out cameras with Foveon's tech. With competition, we might see further refinement of the design.
Here's a comprehensive review of Sigma's camera.
Actually, it is important to match the quality of the lens to the resolution of the CCD. Too fine a lens will cause aliasing artifacts. This can be understood by the Nyquist Theorem.
Once an aliased image is captured by the CCD, no amount of image processing can remove the artifacts. That is why high end digital cameras like the Nikon D1 contain an optical low-pass filter between the lens and the CCD that purposefully degrades the quality of the lens assembly.
The most important part of any camera is the lens. There are two main problems with lenses. Chromatic aberration causes colour fringing due to the focal length of the lens being different at different wavelengths. It can be corrected by using compound lenses {one positively-dispersing lens and one negatively-dispersing lens} or low-dispersivity materials. Spherical aberration causes distortion of the image due to the lens surface not being perfectly spherical, and thus the focal length varying over the surface. It can only be corrected by grinding lenses well.
A bad lens will produce a bad image regardless of the image sensor. Sometimes an image sensor will not have enough resolution to detect the distortion due to chromatic and spherical aberrations. But when the same manufacturer slaps a new sensor on last year's lens, the new sensor can pick up better on the aberration and the pictures end up looking lousy.
Another feature to bear in mind is hardware {optical} zoom. Don't buy a camera without it and don't reject a camera for not having software {digital} zoom -- your favourite graphics editor can do this for you.
Cheap image sensors are invariably noisy. Big pixels can hold more initial charge, therefore can accept more light in the course of an exposure. The sensor will only be saturated in really bright light, and the amount of charge remaining on the pixel {which is a measure of how much light didn't hit it} can be measured more accurately: one "unit" on the ubiquitous 0-255 scale represents many electrons. But more silicon costs more money. Small pixels don't have the same capacitance, so can't accept as much light before becoming saturated -- you have to run a shorter exposure. And the number of electrons per ADC count is smaller. The net result of having a higher density in the image sensor is that even in bright light, the resulting pictures will look a little bit as though they were taken in poor light. Of course, you can remove the noise by downsampling, but then you lose the benefit of the higher-res sensor.
And what's with the confusing term "digital SLR" ? As far as I can see, all digital cameras with LCD viewfinders are by definition SLRs, since the same lens is used for viewing and taking the picture.
Je fume. Tu fumes. Nous fûmes!
The article seems to be making the argument that a smaller format sensor won't be as sensitive as a larger sensor, but I'm not sure I buy this.
A smaller sensor is more noisy and more prone to chromatic abberation. Which is why my old EOS D30 with a large 3MP CMOS sensor produces better pictures than Sony's F828, which crams 8MP onto a tiny CCD. 3MP prints great up to 9x6" and is uable at 12x8". It's difficult to get a good print off a CCD camera above 7x5". Larger images don't need to be distorted as much by the lens to be focussed down onto a larger sensor, and that matters. More photons per unit area matters for faithful colour reproduction.
But like another poster said, most of these images are destined to be viewed only on screen, so most of the resolution is wasted. About the only thing it's useful for is giving the freedom to crop.
Well, it's fact. The larger the surface area of each cell, the better signal to noise ratio you will get. CMOS yields better quality than CCD, as well- although the margin has dropped as CCD sensors and the electronics behind them have improved faster(due to everyone and their grandmother working with CCDR sensors) than CMOS.
This phenomenon can be seen clearly in both the non-CMOS 14 megapixel Kodak 14n, or the Sony F828, which has a VERY tiny 8 megapixel CCD sensor. Both are horrendously noisy at their lowest ISO settings.
My Canon 10D has better noise characteristics at about 400 ISO than my Canon G1 had at 50 ISO, and 400 is about the limit I feel is appropriate for an 8x10. For images resized to 800x600 for, say, large images linked off a website, ISO 800 or 1600 still yields pretty decent images. The example he gives of buckets of water is flawed, since falling rain isn't *focused* like light is. Light entering a lens is just being focused on a smaller area. Sure the area is smaller, but it's also brighter.
Light is focused, but it's also made up of particles. Further, the smaller the sensor, the smaller the lens. The smaller the lens, the less light is gathered.
Smaller sensors also require much more precise optics and focusing systems(or smaller apertures, limiting light input even further). Tiny sensors are also very prone to flare.
Please help metamoderate.
1. Image quality will be determined by the combination of how many pixels you capture (megapixel count/resolution) and the size of those pixels (sensor size/photosite size), with the weight of the influence going to the photosite size. Thus, resolution being equal between two cameras, the camera with the larger sensor size will give you higher quality captures.
2. Garbage In, Garbage Out applies to cameras too. This should be obvious. Make sure your lens is able to capture all the data you want to feed to your sensor. If you have a full-frame, 24x36mm 11mp Canon 1Ds (the current professional favorite, myself included), you are wasting it's resolution by putting a cheap lens on it. I've noticed, in fact, that even the highest-quality lenses tend to be unable to deliver enough detail to this stunning sensor, so a cheapo lens is going to f*ck you.
3. For professional use, film is now dead. Game over. I've done the head-to-head comparisons. I own medium and large-format cameras. I own a high-end drumscanner. I own a large-format printer. I've compared the quality from my previous breadwinning equipment (medium format film scanned by drumscanner) to my current breadwinning equipment (full-frame digital Canon 1Ds) and the digital kicks film ass. That's why it's my current breadwinner.
Seriously, I had 4x6 foot prints made (notice I said FEET, not INCHES) from drum-scanned 6x7cm transparencies, and from 11mp Canon 1Ds captures, and my own lab couldn't tell the difference. Bye-bye film. And the $10,000 price tag was paid for in film/processing savings before I even got the credit card bill. (for more about how cost affects quality, see below, #5)
4. The best camera for you is all about what you intend to do with it. A camera is just a tool. Pick the right one for the job. Because of this, most professionals have, on average, more than 3 different camera systems. So, decide what you want the camera for, and the rest of the decisions about it's suitability get easy.
The most important factor is usually not sheer resolution and image quality. It's about usability of design and ease of handling. If it were all about resolution then most photographers would be using 8x10-inch view cameras. But we realize that a stunning, mega-high-resolution image is useless if the important moment we wanted to capture was missed due to slow camera operation.
That's why most pros use medium format or 35mm, and most ams use point'n'shoots.
So, pick a camera that feels good, is understandable to operate, and doesn't get in your way. After these criteria are satisfied THEN you look at resolution/sensor size.
5. The single most important equation for making better photographs is (forethought x volume of action). In other words, think about what you want to achieve with your images, then shoot as much as you can, and hone your results. This is really where digital capture shifts paradigms. Once you go digital, ANY digital, your visual experiments cost you nothing.
With film, every time you want to try something new, you are still paying for film and processing (even if you own your own darkroom). This means, effectively, that film and processing are an economic tax on your creative growth.
So, as long as you stay focused on what you want to achieve (rather then just shooting because you can), buying ANY decent digital camera will yield you better results then sticking with film, and it's use tax.
Class dismissed.
He also said there was some thru the mail company, seattle film, or something like that. they would send people film, you send the film to them for processing. The quality on the film sucked because it was some different technology, and that you were locked into their scheme because you couldn't get it developed anywhere else..
yep. they basically sold 35mm movie film in 24 frame strips. movie film doesn't have the same high quality requirements as still photography film because any problem in a frame is corrected 1/24th of a second later.
movie film is therefore much much cheaper per frame than good photographic film. so they were making out like bandits when they hooked someone. and because it doesn't use e-6/c-41 chemicals you had to get it developed either at a motion film lab (not likely) or with them.
btw, movie film also has a really short shelf life unless kept in special volts at exact temperatures. this is true even AFTER the film is developed!
seattle filmworks was one a very nasty scam for several decades. a few years ago they finally switched to (really crappy) c-41 film.
in this age of communication i'm just not getting through
Too fine a lens will cause aliasing artifacts. This can be understood by the Nyquist Theorem.
This depends on the shape of the CCD active areas that are used to capture photons. In the "worst case" where the receptors are essentially discrete points on a grid, an optical blur is needed so photons that would otherwise land in between the sensors have a chance to be captured. In practice, I would guess that the sensors cover about 50% of the usable area, so the remaining 50% must be made up with low-pass filtering to avoid aliasing. (Think of filming headlights; if they're in focus they'll be two discrete points of light, but as you defocus the lights will expand until they overlap.)
A similar problem also comes up in motion video; the aperture is typically open 75% of the time, then closed 25% while the film advances. This results in motion aliasing such as helicopter blades and wagon wheels spinning backwards, etc. Digital video may be able to substantially reduce this problem, but ironically most people have grown accustomed to it, to the point where non-aliased video simply doesn't "look right."
The Foveon approach is a step in the right direction for image capture, since the Bayer interpolation from most other cameras is prone to all sorts of artifacts. Perhaps a camera could be built that would expose the same CCD array through red, green and blue filters in sequence, then apply software to compensate for slight motion between frames.
Similarly, imagine a camera that would expose the CCD for 1/10,000 of a second, then 1/1000, then 1/100, then 1/10, and combine the resulting frames into a single high-dynamic-range image. When the sun is millions of times brighter than the shadows, [0..255] simply isn't going to do justice.
In my opionion, the next few years of digital photography is going to be mighty interesting.
Weeks of coding saves hours of planning.