If this is your first visit, be sure to check out the FAQ by clicking the link above. You may have to register before you can post: click the register link above to proceed. To start viewing messages, select the forum that you want to visit from the selection below. |
|
|
Thread Tools | Display Modes |
#1
|
|||
|
|||
Resolution limit of image sensor
Hi NG,
Can someone please explain to me if there is a connection between the Nyquist sampling theorem and the resolution limit of a digital image sensor? I mean, does it imply something like a lowest mark as far as pixel spacing is concerned? - I´m quite new to digital photography and keep reading about this stuff but must admit that it´s by far too theoretical for me! Marc Wossner |
#2
|
|||
|
|||
Resolution limit of image sensor
I would assume that sampling theorema apply to post processing (A/D +
following processing) but not to the image sensor itself where quantum efficiency plays a role. Some transfer functions can be used to characterize lenses. "Marc Wossner" schrieb im Newsbeitrag ups.com... Hi NG, Can someone please explain to me if there is a connection between the Nyquist sampling theorem and the resolution limit of a digital image sensor? I mean, does it imply something like a lowest mark as far as pixel spacing is concerned? - I´m quite new to digital photography and keep reading about this stuff but must admit that it´s by far too theoretical for me! Marc Wossner |
#3
|
|||
|
|||
Resolution limit of image sensor
In article . com "Marc Wossner" writes:
$Can someone please explain to me if there is a connection between the $Nyquist sampling theorem and the resolution limit of a digital image $sensor? Yes, there's a connection, but it's not the only factor. My Canon EOS 20D's sensor has a resolution of 3504x2336 (that's effective pixels; as with most sensors, there are some additional pixels that don't actually get used as part of the image data). Nyquist tells us that this can produce at most half that many line pairs. That sets an upper limit on resolution for any given sensor. But there are other factors that come into play. Lenses are not perfect; they all result in some level of loss of contrast and/or sharpness. So if you were to quadruple the number of pixels by doubling the number in each dimension, that wouldn't necessarily result in images with twice the sharpness or twice the detail, if for no other reason than that the lens may not be up to the task. You could also take the same number of pixels and make them larger. The 20D's sensor is about 22x15mm, so it has approximately 160 pixels per millimeter. The 1D Mark IIN has the same number of effective pixels but in a ~29x19mm sensor, yielding about 120 pixels per millimeter. So despite the same number of pixels, a lens that deliver sharper results at 60 lp/mm than at 80 lp/mm will give you sharper images on the 1D IIN than on the 20D. $ I mean, does it imply something like a lowest mark as far as $pixel spacing is concerned? You can certainly increase the maximum resolution that the sensor can capture by making the pixels smaller. But then you run into other problems. Noise is one major problem here. There's a certain level of background noise. A larger pixel captures more photons, so the ratio of signal (photons) to noise can be pretty good. A smaller pixel captures fewer photons, yielding a lower signal to noise ratio. There's also the issue of Poisson distribution; the arrival and distribution of photons are random, and even if you take a picture of a subject which is perfectly even, some pixels will receive a few more photons than others. Again, in a large pixel, this random variation is small relative to the total number of photons, while in a small pixel, this variation can be significant. If you've ever compared a typical shot at ISO 400 from a compact digital P&S (which has a relatively small sensor and therefore tiny pixels) to a typical shot at ISO 400 from a DSLR (which has a relatively large sensor and therefore large pixels), you'll understand this in practical terms: the P&S picture is significantly noisier than the DSLR picture. There are also issues about data volumes as the number of pixels rises. An 8 MP camera usually produces JPEGs that are in the 3-4 MB ballpark. A 16 MP camera would produce files that are about twice that size. How big a file do you need? How big a file do you want to have to store? How much flash memory do you want to have to take on holiday with you in order to store all your pictures? And speed ... the 1D IIN can take about 8.5 frames per second, with a resolution of 8.2 MP. The 1Ds II has a 16.7 MP sensor and can only shoot at about 4 frames per second. It's not because the mechanical bits can't keep up (both cameras are very similar mechanically, and are based on a film camera, the 1V, which can get up to 10 fps) or because Canon's engineers got lazy when designing the 1Ds II; there's simply too much data to be moved around. There are digital backs for medium-format cameras which yield tens of megapixels, and they typically can't even do two frames per second, for the same reason. -- Stephen M. Dunn ---------------- http://www.stevedunn.ca/ ---------------- ------------------------------------------------------------------ Say hi to my cat -- http://www.stevedunn.ca/photos/toby/ |
#4
|
|||
|
|||
Resolution limit of image sensor
"Marc Wossner" wrote in message
ups.com... Hi NG, Can someone please explain to me if there is a connection between the Nyquist sampling theorem and the resolution limit of a digital image sensor? In imaging terms, Nyquist defines maximum possible sharpness in terms of the spacing between pixels. More pixels equals more sharpness, all else being equal. I mean, does it imply something like a lowest mark as far as pixel spacing is concerned? Another way to think of it is that - optics aside - the most detail you can have in an image is every other pixel being black and white, in a checkerboard pattern. If you attempt to represent finer detail, your image gets killed by a crazed pattern called moire. I´m quite new to digital photography and keep reading about this stuff but must admit that it´s by far too theoretical for me! There's a world of mathematical theory behind sharpness: Fourier Transforms, Modulation Transfer Functions. Some of propeller head types us love this, but it's a world of hurt if you go in there unwillingly. The rest of us can ignore it because the following rules of thumb will get the job done, and only fourth grade arithmetic is required to apply these rules of thumb to your own images. Rules of thumb: A razor sharp print is about 320 pixels per inch. Some people claim they can see more than that - hah. A very acceptably sharp print is about 200 pixels per inch. At 100 pixels per inch, ordinary people may be aware of some jagginess in the image. Less than 100 pixels per inch is generally considered unacceptable but look how great your monitor looks at even less than 100 pixels per inch. I've gotten away with a 72 pixel per inch (also know as dot per inch in the trade) it on a few occasions. Viewing distance matters, so add to this the fact that people stand back more from a larger print, and you can go even larger than the ppi numbers would indicate. Check this out yourself by walking up to a billboard - it's lucky to be one pixel per inch. Arithmetic: You can ignore megapixels. All we need is the image dimensions in pixels. My older camera took (and still takes) a 2048 by 1536 image. Dividing by 320 says a razor sharp image from this camera would be about 6 by 4 inches. Dividing by 200 gives about a 10 by 8 image. Larger than that, and the image starts to look a bit fuzzy, though remember that fuzziness is more acceptable in a larger image. My newer camera takes a 3204 by 2136 image. Dividing by 320 says that I can print a razor sharp 8x10 image. Dividing by 200 gives me a pretty darn sharp 16 by 10 image, and a very nice 19 by 11. In fact, this camera produces 11 by 19 prints that look wonderful at just under 200 ppi. BTW - curvemeister class starts tomorrow, Sunday the 7th: http://www.curvemeister.com/support/class/index.htm -- Mike Russell www.curvemeister.com/forum/ |
#5
|
|||
|
|||
Resolution limit of image sensor
"Marc Wossner" wrote in message ups.com... Hi NG, Can someone please explain to me if there is a connection between the Nyquist sampling theorem and the resolution limit of a digital image sensor? I mean, does it imply something like a lowest mark as far as pixel spacing is concerned? - I´m quite new to digital photography and keep reading about this stuff but must admit that it´s by far too theoretical for me! As the image detail approaches one-half the spatial sampling frequency, aliasing starts to become a problem. Aliasing means that artifacts show up that were not in the scene but were caused by too low of a spatial sampling frequency or too much image detail. The fix is a blur filter (or anti-alias filter) mounted on top of the sensor. |
#6
|
|||
|
|||
Resolution limit of image sensor
Thank you all for your exhaustive replies, I understand things much better now! So, as I still use silverfilm too, can it be said that the Nyquist theorem limits it´s resolution in a very fundamental way as well but is lower because the silver halides are smaller than the individual pixels? Marc Wossner |
#7
|
|||
|
|||
Resolution limit of image sensor
Just to check my understanding: If I have a sensor with 3034 horizontal
pixels and a spacing of 7,8 µm it can resolve: max frequency = scan frequency/2 = 1517 lines in this direction = structures wich are not closer than 3,9 µm to each other Is that correct? Marc Wossner |
#8
|
|||
|
|||
Resolution limit of image sensor
Marc Wossner wrote: Hi NG, Can someone please explain to me if there is a connection between the Nyquist sampling theorem and the resolution limit of a digital image sensor? I mean, does it imply something like a lowest mark as far as pixel spacing is concerned? - I´m quite new to digital photography and keep reading about this stuff but must admit that it´s by far too theoretical for me! Marc Wossner There was much work done on this in the thirties by early TV engineers, since in the vertical direction image tube TV and kinescopes are "sampled" systems. Much of the work was done' experimentally, and the guru was an engineer by the name of Ray Kell. The resulting widely used value, now called the Kell factor, was around 0.7. That is, for a system with N samples in a given direction (either vertical or horizontal) one can resolve about 0.7N lines. When we got our hands on our first CCD chip at work in the late '70s, I did an analysis (numerical, sort of Monte Carlo) and found a value very close to that for mosaic arrays. While it did depend a bit on fill factor, the dependence wasn't strong. I still use 70% as a good expectation. |
#9
|
|||
|
|||
Resolution limit of image sensor
In article .com "Marc Wossner" writes:
$So, as I still use silverfilm too, can it be said that the Nyquist $theorem limits it=B4s resolution in a very fundamental way as well but $is lower because the silver halides are smaller than the individual $pixels? You'll need to find someone who understands film well for a solid answer on this one. But in any system that uses discrete samples, Nyquist applies in some fashion. A given silver halide crystal can't represent both black and white simultaneously, so you need two of them side by side, one black and one white, and bingo, there's Nyquist at work again. But film is more complex than that. On a typical digital sensor, the sensor sites are all the same size, and are laid out in a perfectly regular pattern in which they do not overlap. Not so with film; the crystals are not all the same size, and they are not laid out regularly, and according to the diagrams I've seen in film technical data sheets, an emulsion layer is more than one crystal thick. (For an example, look at Fuji's datasheet for Velvia 50: http://www.fujifilmusa.com/JSP/fuji/...AF3-960E_1.pdf and go to page 6.) -- Stephen M. Dunn ---------------- http://www.stevedunn.ca/ ---------------- ------------------------------------------------------------------ Say hi to my cat -- http://www.stevedunn.ca/photos/toby/ |
#10
|
|||
|
|||
Resolution limit of image sensor
Lionel wrote:
On Sat, 6 Jan 2007 12:34:57 -0800, "Mike Russell" -MOVE wrote: There's a world of mathematical theory behind sharpness: Fourier Transforms, Modulation Transfer Functions. Some of propeller head types us love this, but it's a world of hurt if you go in there unwillingly. Nicely put. My reaction, too. Rules of thumb: A razor sharp print is about 320 pixels per inch. Some people claim they can see more than that - hah. A very acceptably sharp print is about 200 pixels per inch. At 100 pixels per inch, ordinary people may be aware of some jagginess in the image. Indeed. My eyesight is very good, & I can't tell the difference between 250 & 300DPI. I can just tell the difference between 200 & 300 DPI - *if* I hold both prints up to my nose in sunlight - but 200 DPI is plenty sharp enough in every other situation. Er, Lionel, *ppi*! And I know you know the difference. IAE, it's nice to see you back posting here on photography. Now, who can tell the diff between prints printed at 720 dpi vs. 1440 vs. 2880 dpi? -- John McWilliams I know that you believe you understood what you think I said, but I'm not sure you realize that what you heard is not what I meant. |
Thread Tools | |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
SENSOR SIZE AND RESOLUTION | Wladyslaw Wojciechowski | Digital Point & Shoot Cameras | 1 | November 21st 06 10:29 AM |
SENSOR SIZE AND RESOLUTION | Wladyslaw Wojciechowski | Digital ZLR Cameras | 3 | November 20th 06 08:45 PM |
given file size limit, optimize jpeg level and resolution? | peter | Digital Photography | 8 | May 25th 06 01:01 AM |
Sensor resolution, any sites with actual measurements? | Rich | Digital Photography | 19 | March 8th 06 08:39 AM |
sensor size vs resolution | freightcar | Digital Photography | 9 | February 5th 06 02:40 AM |