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 |
#31
|
|||
|
|||
Math question - sort of
John Navas wrote:
On Sun, 27 Sep 2009 21:02:51 -0400, Alan Browne wrote in : You Are The Weakest Link wrote: On Sun, 27 Sep 2009 10:55:01 -0400, Alan Browne wrote: John Sheehy wrote: John Navas wrote in : On Fri, 25 Sep 2009 17:10:27 -0400, "Charles" wrote in : Let the reader decide: http://www.luminous-landscape.com/tu...solution.shtml Sensors for larger formats are approaching the diffraction limit of real lenses, and it is more difficult to get high levels of aberration suppression for them. The point is that you cannot fully exploit the resolution potential of high-resolution sensors with regular mass-produced lenses, particularly for larger formats. The lenses are to blame for any optical issues with high densities. The higher density *NEVER* exacerbates any lens problems. Lower densities lower the resolution, so you see less of everything, including subject detail. You position is all "talk" and "logic". You can not demonstrate what you believe, because it only exists in bad logic and bad paradigms. Here's what happens when you try to demonstrate, and go about it the right way: You shoot the same scene with the same lens, same ISO, same Av and Tv, and then you use a converter with no noise reduction, and upsample critical crops from both images to the same subject size. No matter how much lens fault is brought into the light with the higher density, the higher density still has a more accurate rendition of the subject, because those faults ARE ALWAYS THERE, REGARDLESS OF PIXEL DENSITY. Less agressive sampling does not avoid lens issues; it just makes it harder to tell why the image has so much less real subject detail. Is that another way of saying the Kodak empirical formula for end image resolution (on film) is... 1/sqrt(res_out) = 1/sqrt(res_lens) + 1/sqrt(res_sensor) ? So increasing either the sensor density or the lens resolution results in higher output resolution, though of course with diminishing returns. It's not an "either/or" venture. It's an "and" issue. You really don't know how to read and understand that "increasing either" also includes "increasing both" do you? You manufacture resolution. You are always limited by the resolution of the lens. Increasing resolution of the sensor past that point just copies lens artifacts more faithfully. It's not a choke and it's not that simple - that is where the empirical formula from Kodak comes from. To be sure, the returns are diminishing on both sides. |
#32
|
|||
|
|||
Math question - sort of
You Are The Weakest Link wrote:
On Sun, 27 Sep 2009 21:02:51 -0400, Alan Browne You really don't know how to read and understand that "increasing either" also includes "increasing both" do you? You don't know how to comprehend that increasing either does NOT include increasing both. OR != XOR. Look it up. Write a logic table. -Wolfgang |
#33
|
|||
|
|||
Math question - sort of
John Navas wrote in
news On Mon, 28 Sep 2009 00:27:28 GMT, John Sheehy wrote in : I meant the density itself. Of course, microlenses could be poorly designed. Even then, however, oversampling allows extemely easy and smooth correction of CA, both from the lens, and that generated by poor microlenses. Oversampling does not facilitate correction of CA Yes, it does. You lose all your eggs when you put them in big pixels and have to shift them by non-integer numbers of pixels. or microlens aberrations. That's all in the design - a microlens doesn't have to cause any CA-like effects of its own. |
#34
|
|||
|
|||
Math question - sort of
John Navas wrote in
: On Tue, 06 Oct 2009 20:59:01 -0500, John Sheehy wrote in : John Navas wrote in news Oversampling does not facilitate correction of CA Yes, it does. You lose all your eggs when you put them in big pixels and have to shift them by non-integer numbers of pixels. Not so. So what do you do, use point-sampling and wind up with a nearest-neighbor jaggy mess? |
#35
|
|||
|
|||
Math question - sort of
John Sheehy wrote:
John Navas wrote in : On Tue, 06 Oct 2009 20:59:01 -0500, John Sheehy wrote in : John Navas wrote in news Oversampling does not facilitate correction of CA Yes, it does. You lose all your eggs when you put them in big pixels and have to shift them by non-integer numbers of pixels. Not so. Although it does make it easier. So what do you do, use point-sampling and wind up with a nearest-neighbor jaggy mess? There are resampling methods derived from radio astronomy that can handle this situation accurately but they are computationally expensive. Bilinear spline is about the cheapest half decent option found in standard packages. But there are better ones if you have resources to burn. It ends up with the law of diminishing returns so how far you push it is really determined by how unique or irreplaceable the image is. If you have the option then oversampling the measured data by about 1.5x the Nyquist theoretical minimum for a monochrome imaging system is worthwhile. Otherwise you may see obvious jaggies in the raw image. Beyond that you are not gaining much although for a Bayer sensor you still get a bit of extra chroma information out to 2x oversampled. Regards, Martin Brown |
#36
|
|||
|
|||
Math question - sort of
John Navas wrote:
On Wed, 07 Oct 2009 22:14:35 +0100, Martin Brown wrote in : There are resampling methods derived from radio astronomy that can handle this situation accurately but they are computationally expensive. Bilinear spline is about the cheapest half decent option found in standard packages. But there are better ones if you have resources to burn. It ends up with the law of diminishing returns so how far you push it is really determined by how unique or irreplaceable the image is. If you have the option then oversampling the measured data by about 1.5x the Nyquist theoretical minimum for a monochrome imaging system is worthwhile. Otherwise you may see obvious jaggies in the raw image. Beyond that you are not gaining much although for a Bayer sensor you still get a bit of extra chroma information out to 2x oversampled. There's a big difference between sampling the same signal multiple times (time) and breaking up a photosite into multiple photosites (area). Actually there isn't all that much of a difference apart from the obvious one that a time series is one dimensional and so a lot more amenable to analytical techniques when sampled at equal intervals. Time sampled data is usually integrated over a time delta-t rather than a true snapshot of the signal by a flash converter at exact time t. If, for example, you have a single point photosite at or above the resolution limit of the lens, and you divide it into four photosites, you are not adding resolution (limited by the lens), you are at most more accurately sampling the luminosity of that point, depending on the tradeoff between photosite size (sensitivity) and the noise floor. Indeed, but having some of that extra data can make post processing deconvolution more reliable provided that you have not traded signal to noise. The point here is that an undersampled digital image does present some difficulties for post processing to remove chromatic and other abberations. They are not insurmountable but it is easier with an oversampled image. Regards, Martin Brown |
#37
|
|||
|
|||
Math question - sort of
In article , John Navas
wrote: Theory is interesting, but in the real world it's just not an issue -- Panasonic has long been correcting CA with in-camera processing and without oversampling, and the result is excellent (not a "nearest-neighbor jaggy mess"). Are you still using that Panasonic piece of ****? I guess it doesn't matter since you don't have the ability to create a decent image. |
#38
|
|||
|
|||
Math question - sort of
On 10/8/09 10:48 PM, in article , "Mr. Strat" wrote: In article , John Navas wrote: Theory is interesting, but in the real world it's just not an issue -- Panasonic has long been correcting CA with in-camera processing and without oversampling, and the result is excellent (not a "nearest-neighbor jaggy mess"). Are you still using that Panasonic piece of ****? I guess it doesn't matter since you don't have the ability to create a decent image. And he's still using the Motorola phone charger... |
Thread Tools | |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
a techie sort of question about p&s cameras and optical viewfinders | albert | Digital Photography | 9 | December 15th 08 08:03 PM |
A meter math question | Steven Woody | 35mm Photo Equipment | 7 | April 11th 07 04:51 PM |
help with aperture math | Beach Bum | Digital SLR Cameras | 18 | February 15th 06 03:18 AM |