Forgot Password
Pentax Camera Forums Home
 

Reply
Show Printable Version Search this Thread
03-08-2009, 02:35 PM   #76
Senior Member




Join Date: Feb 2008
Location: madison
Posts: 239
I think he is teaching us the importance of SR/IS.

QuoteOriginally posted by *isteve Quote
Sorry I see nothing very impressive in the picture you linked to. Am I missing something?



Last edited by cousinsane; 03-08-2009 at 05:22 PM.
03-08-2009, 05:08 PM   #77
Veteran Member
falconeye's Avatar

Join Date: Jan 2008
Location: Munich, Alps, Germany
Photos: Gallery
Posts: 6,863
QuoteOriginally posted by jeffkrol Quote
image-based autofocus will outpace phase-detect autofocus in the future, but not necessarily soon, Watanabe said.
Thanks for the quote.
This is exactly what I think. I eagerly await contrast AF for satisfactory accuracy. It is just not fast enough (yet).

Having developped a fast contrast AF algorithm (for microscopy) myself, do you think it is possible to try it out in a DSLR? Is there a DSLR with an open source firmware open enough to steer the AF from captured life view? And programmable in C? I am aware of CHDK (for Canon P&S) but that doesn't seem to support this level of control.
03-08-2009, 06:07 PM   #78
Veteran Member




Join Date: Mar 2007
Location: Philippines
Posts: 1,399
QuoteOriginally posted by falconeye Quote
Thanks for the quote.
This is exactly what I think. I eagerly await contrast AF for satisfactory accuracy. It is just not fast enough (yet).

Having developped a fast contrast AF algorithm (for microscopy) myself, do you think it is possible to try it out in a DSLR? Is there a DSLR with an open source firmware open enough to steer the AF from captured life view? And programmable in C? I am aware of CHDK (for Canon P&S) but that doesn't seem to support this level of control.
I haven't seen any DSLR who has had its firmware hacked well enough to be able to put in custom codes for AF algorithms. It would be nice if a company offered one, though, even if they put it the precondition that the warranty will be void if one modifies the factory firmware.
03-08-2009, 07:34 PM   #79
Veteran Member
jeffkrol's Avatar

Join Date: Sep 2006
Location: Wisconsin USA
Photos: Gallery
Posts: 8,435
QuoteOriginally posted by falconeye Quote
Thanks for the quote.
This is exactly what I think. I eagerly await contrast AF for satisfactory accuracy. It is just not fast enough (yet).

Having developped a fast contrast AF algorithm (for microscopy) myself, do you think it is possible to try it out in a DSLR? Is there a DSLR with an open source firmware open enough to steer the AF from captured life view? And programmable in C? I am aware of CHDK (for Canon P&S) but that doesn't seem to support this level of control.
though it wouldn't be color, how about a kit?
Audine: Table of Contents

03-08-2009, 07:38 PM   #80
Veteran Member




Join Date: Jun 2008
Location: Guatemala
Posts: 338
I've always wanted a company to come out with a SLR with open source firmware, so much would be doable. If only I knew more about reverse engineering and low level programming, I would try my hand at making something for my GX-10.
03-08-2009, 11:24 PM   #81
Veteran Member




Join Date: Nov 2006
Location: Hong Kong
Posts: 1,934
The Truth Is..

QuoteOriginally posted by cateto Quote
Akira Watanabe, head of dSLR development in Olympus, declares in an interview for CNET News that "12 megapixels is enough".
12MPs is just more than enough for such a small sensor of them without having more noise that they could not manage!

The grape at other lands that Oly cannot get to taste must be sour!
03-09-2009, 12:19 AM   #82
jay
Inactive Account




Join Date: Feb 2009
Location: Lincoln, Nebraska, USA
Posts: 65
Holy smokes! Look at all these people that don't know what they're talking about!

It all started with this:

QuoteOriginally posted by er1kksen Quote
Good god people, how many times are we going to have to go over the fact that higher pixel density DOES NOT degrade image quality/high ISO performance?
It's understandable -- this can be very confusing, because there's two different pictures going on

Many of us understand the particle model. A pixel is a square light-sensitive detector. Obviously, the larger the pixel is, the more light it can collect, which means that for a given exposure, a larger sensor means a brighter image.

In this model, light falling on a sensor is like rain falling into a rain gauge. The bigger the gauge, the more water you get.

But by looking at the electromagnetic wave model -- which is equally valid -- things aren't as simple. Light oscilates at different wavelengths, and each pixel is an antenna. The size of the photosite determines the transfer characteristics. When I say "transfer characteristics," I don't mean a simple linear model showing that as the sensor gets bigger or smaller, light is easier or harder to collect. It's not that simple.

In general, all things being equal, when you increase the pixel density of a sensor, you're shrinking each photosite, which redueces the amount of light, and increases the noise of the photo.

And, to add insult to injury, more megapixels doesn't mean sharper images. There's really no correlation at all once you hit the dozen-or-so megapixel mark on a APS-C sensor. At that point, the transfer characteristics of the photosite becomes less intuitive.

There's other issues, too, when we're talking about image quality and sensors "outresolving lenses" (always a loaded statement to make). Sampling has to do a lot with what I was talking at previously, and I don't have time to explaiin it here, however, there was a great article on LL:

Do Sensors “Outresolve” Lenses?

Check it out.
03-09-2009, 02:00 PM   #83
Senior Member




Join Date: Jun 2008
Location: Atlanta, Georgia
Photos: Gallery
Posts: 182
QuoteOriginally posted by brkl Quote
That doesn't look all that sharp to me compared to digital. Here's jin's excellent shot from the Post your photos! forum, this thread, along with a crop of a similar area as the crop of that BW photo. Tell me jin's shot doesn't have oodles more image information, and that's nowhere near 35mp.
While it is an very good example of the excellent quality that can be acheived with K20D (and more importantly a very good photo from an artistic point of view), I can not agree that resolution wise Jin's picture contains a lot more information. The size of the crops (i.e. how large part of the picture the crop contains) are very different and it is therefore not easy to compare directly. I tend to more agree with falconeye in that looking at absolute resolution of the K20D sensor in term of lp/mm and the example of the film used in the example, they are similar. Thus scaling the same pixel density as K20D to a FF sensor (x2.25), the example in the link would be equal to about 32 Megapixels on a FF camera.

Best regards,
Haakan

03-09-2009, 03:47 PM   #84
Veteran Member
jeffkrol's Avatar

Join Date: Sep 2006
Location: Wisconsin USA
Photos: Gallery
Posts: 8,435
QuoteOriginally posted by jay Quote
Holy smokes! Look at all these people that don't know what they're talking about!



Do Sensors “Outresolve” Lenses?

Check it out.
You are sooo right. Think you should move your debate to here:
Re: Very little: News Discussion Forum: Digital Photography Review
But never mind. Yes if you make the pixel four times bigger, and scale the gate you end up with four times the capacitance, but you can only fit one quarter of the number of pixels in the same space, so the saturation electron density, and thus the DR and noise remains the same.
> If you double the feature size where x and y = 2; (x*y)/(x*y + x + y
> + 1)
> = 44% fill factor. Basically each pixel takes 3x by 3y. If you have a
> 100 unit by 100 unit sensor, it will be a 1089 pixel sensor.

As I said, it all scales, including the 'space between pixels' so your argument is simply wrong. That's an objective fact.
> Each pixel in the 1089 pixel design will have twice the dynamic range
> as the 2500 pixel sensor.

Except that your argument is fallacious, so in (objective) fact the DR will be independent of pixel density, as tests indicate.
> Let's pretend this sensor size is at the
> limits of diffraction at 250 pixels.

Let's not. Diffraction is another canard raised by those who don't understand it properly.
> The 2500 pixel sensor will gain
> no resolution advantage over the 1089 pixel sensor, yet will have
> half the dynamic range.



MORE on this and a refute of the luminous landscape chart...........
Here is a web page which includes the wrong diffraction limit you cite:
http://www.luminous-landscape.com/tutorials/resolution.shtml

It is wrong for two reasons:

It states that while the Rayleigh criteria is adequate for resolving lines with the human eye a sensor needs more contrast; but the examples given don't support this. In fact they are wrong, sharpening can be used to increase the high frequency contrast so the Rayleigh criteria is correct for sensors as well as your eyes.

They then go on to claim that while a Bayer sensor without an antialiasing filter needs more pixels for a given resolution; that one with an antialiasing filter has the exact same resolution limit as a monochrome sensor which is absurd.

If you want a pixel count which is approximately correct for a given aperture you can use their table 2 column 3 (for Bayer sensors) for green light. That shows 373 pixels per millimeter at f/8 which is 47 megapixels for an APS-C sensor. I expect that will be the approximate practical pixel limit.

http://forums.dpreview.com/forums/read.asp?forum=1019&message=31264423
http://forums.dpreview.com/forums/postersprofile.asp?poster=hiixiwilhvij
Bottom line APS-C has a way to go....
before you throw noise in this keep in mind:
> ii) Once again, the quoting of 'laws of physics' that don't exist.
> Increasing pixel density does not increase noise. Write it out 100
> times. John understands it, he has been through many lengthy
> discussions where his opinions have been tested to breaking point,
> and come through. I don't think you have, you've just picked up the
> easy to believe myth and swallowed it whole, uncritically.

Well, this was discussed at length and I think John would agree that read noise does set a practical lower limit on pixel size. Remember that read noise does not go down with pixel size: pixel capacitance is set by the floating diffusion size of the read-out transistor *not* the pixel active area. There are ways to reduce read noise and the latest sensors (like the one in the 5D2) are the best yet; but in practice read noise is unlikely to be so small that it can be ignored any time soon.
http://forums.dpreview.com/forums/read.asp?forum=1018&message=31264707

Last edited by jeffkrol; 03-10-2009 at 03:50 PM.
03-16-2009, 06:01 PM   #85
Veteran Member
jeffkrol's Avatar

Join Date: Sep 2006
Location: Wisconsin USA
Photos: Gallery
Posts: 8,435
More diffraction "stuff"

Found this at the 12mp ect. thread........
Just some more food for thought for those that think we reached any "limit"
Diffraction and sharpening: Open Talk Forum: Digital Photography Review

The reason that the traditional diffraction limits underestimate the number of pixels that you can achieve is that they don't consider the detail that can be recovered by sharpening. In this post: Re: Several reasons...: Open Talk Forum: Digital Photography Review


I explain that the pixel spacing to recover all the available information with diffraction after using sharpening is half the usually cited value. When using sharpening the table below shows the diffraction limited pixel pitch and the resulting full frame mega-pixels based on the f-stop used. Note that in the real world other sources of blurring like the anti-aliasing filter and lens aberrations exist that would suggest that fewer megapixel's are usable; but with Bayer sensors only half the sensor mega-pixels (the green ones) yield the luminance information anyway so a Bayer sensor with the full number of listed pixels would still be reasonable.

fstop microns FF-Mpixels
4.00 : : 1.3 : : 517
5.66 : : 1.8 : : 258
8.00 : : 2.6 : : 129
11.31 : : 3.7 : : 65
16.00 : : 5.2 : : 32
22.63 : : 7.3 : : 16
32.00 : : 10.3 : : 8

The bottom line is that even with a 16 Megapixel full frame sensor f/22 is not past the sensor diffraction limit with sharpening.
03-17-2009, 05:49 AM   #86
Veteran Member
falconeye's Avatar

Join Date: Jan 2008
Location: Munich, Alps, Germany
Photos: Gallery
Posts: 6,863
QuoteOriginally posted by jeffkrol Quote
though it wouldn't be color, how about a kit?
Audine: Table of Contents
Thanks for the link.
Unfortunately, this kit offers nothing more than a web cam w/o the lens today does as well. This is from the old days where a decent CCD camera body suitable for astrophotography was expensive.

Following this idea, I could use a web cam with continous readout to USB and a (slow!) focus motor for astrophotography. Or I use some standard electromotor with controller to drive the Pentax srew-drive of a Pentax lens mounted on an optical bench. All of this goes beyond programming. More than I would be ready to invest right now...
03-17-2009, 06:02 AM   #87
Veteran Member
jeffkrol's Avatar

Join Date: Sep 2006
Location: Wisconsin USA
Photos: Gallery
Posts: 8,435
QuoteOriginally posted by falconeye Quote
Thanks for the link.
Unfortunately, this kit offers nothing more than a web cam w/o the lens today does as well. This is from the old days where a decent CCD camera body suitable for astrophotography was expensive.

Following this idea, I could use a web cam with continous readout to USB and a (slow!) focus motor for astrophotography. Or I use some standard electromotor with controller to drive the Pentax srew-drive of a Pentax lens mounted on an optical bench. All of this goes beyond programming. More than I would be ready to invest right now...
Canon 300D is probably your best chance here. Been hacked a whole bunch.. runs on DOS..........
canondigicamhacking : canondigicamhacking
I should get you the link w/ the microscope imaging "guru". Instead of worrying about exact focus he uses stacked images... LOTS of stacked images...well for the macro stuff at least.
http://www.krebsmicro.com/
http://forums.dpreview.com/forums/read.asp?forum=1018&message=31154182

Last edited by jeffkrol; 03-17-2009 at 06:16 AM.
03-17-2009, 06:32 AM   #88
Veteran Member
falconeye's Avatar

Join Date: Jan 2008
Location: Munich, Alps, Germany
Photos: Gallery
Posts: 6,863
QuoteOriginally posted by jeffkrol Quote
The bottom line is that even with a 16 Megapixel full frame sensor f/22 is not past the sensor diffraction limit with sharpening.
I actually stopped reading and commenting about this topic. Here, I feel invited to complete the picture.

You are right as far as diffraction is concerned. But of course, this is only part of the picture.

These factors play together when it comes to create a decent computational model of image resolution:
- finite sensor resolution
- lens abberation (can be corrected for low noise)
- diffraction (can be corrected up to the Rayleigh limit, sharpening)
- motion blur (can be corrected, deconvolution)
- out-of-focus blur (can be corrected, deconvolution)
- noise (limiting sharpening and corrections, and digital better than film)

I created such a (yet unpublished) computational model taking the first three factors into account.

I followed the tradition to add abberation powers to end up with an overall abberation power [source: from FujiFilm Professional Data Guide, AF3-141E, 2002, p 129]. Resolution is inverse proportional to abberation power.

From my computational model I was able to eliminate finite sensor resolution and diffraction from publihed lens resolution figures and it turned out that the center resolution due to lens abberation (the ray tracing resolution) can be modelled by the following formula:
R(raytracing) = Q * f
f is the f-stop figure (like 5.6) and Q is a lens constant representing a lens optical Quality. Very few best of the best primes have
Q = 120 lp/mm, and very good primes have
Q = 80 - 100 lp/mm (like Pentax FA31)
and so far, I only studied FF SLR primes. Q for zooms would be a lot lower. Beyond 100mm and below 20mm, Q starts to decrease.

So, ray tracing resolution increases with f while diffraction resolution decreases with f. I.e., the higher Q, the lower is the f where resolution maxes out.


Taking all factors into account, I came to the conclusion that 25 MPixels for FourThirds, 40 MPixels for APS-C and 100 MPixels for FF are sort of a sweet spot. A resolution which can be surpassed but where going to the next larger sensor would then be the cheaper solution.
03-17-2009, 06:59 AM   #89
Veteran Member
falconeye's Avatar

Join Date: Jan 2008
Location: Munich, Alps, Germany
Photos: Gallery
Posts: 6,863
QuoteOriginally posted by jeffkrol Quote
Canon 300D is probably your best chance here.
<Untitled>
Thanks again for the links.

Yes, Canon 300D is interesting. Unfortunately, my algorithm needs sort of a live view image to work.

The krebsmicro are fun. I once submitted an image similiar to http://krebsmicro.com/webgal1/File024.html to PPG but it got rejected Image http://krebsmicro.com/webgal1/File022.html is used as a reference on the Helicon website and I recently posted the same image, but stacked using PhotoAcute, in the DOF software thread here.

In order to bring some fun to this thread, let me show a detail from a bee's wing.

Beewing
03-17-2009, 07:04 AM   #90
Veteran Member
jeffkrol's Avatar

Join Date: Sep 2006
Location: Wisconsin USA
Photos: Gallery
Posts: 8,435
QuoteOriginally posted by falconeye Quote




Taking all factors into account, I came to the conclusion that 25 MPixels for FourThirds, 40 MPixels for APS-C and 100 MPixels for FF are sort of a sweet spot. A resolution which can be surpassed but where going to the next larger sensor would then be the cheaper solution.
40MP is still far from the current 14... now how about tripling that but "binning" a triplet (1 filtered w/ each color)..... foveon way but only 2D. You know the only thing preventing this currently is processing power.
Personally I think this will all have to be worked out w/ trial and error.
If I had a slide scanner that went w/ XP I have some mushroom spore photos for you.. I used some cheapo fiber optic stands, a small lens from something and attached it to my Viviatar 550FD. Using a superprogram and it's TTL capability it worked pretty well.. Truck vibration on the street I lived made daytime shots at 1000X impractical
BTW: nicew wing, stack about 50 more shots and bingo....

Last edited by jeffkrol; 03-17-2009 at 07:10 AM.
Reply

Bookmarks
  • Submit Thread to Facebook Facebook
  • Submit Thread to Twitter Twitter
  • Submit Thread to Digg Digg
Tags - Make this thread easier to find by adding keywords to it!
megapixels, pentax news, pentax rumors, world
Thread Tools Search this Thread
Search this Thread:

Advanced Search


Similar Threads
Thread Thread Starter Forum Replies Last Post
The Latest vs My Activity makes no sense to me foxglove Site Suggestions and Help 0 09-01-2010 07:51 AM
Do RML's posts make sense to you? Ash General Talk 46 11-09-2009 11:22 AM
need a back-up telephoto, common sense or old vivitar s1? tvfd911 Pentax SLR Lens Discussion 3 08-13-2009 01:46 PM
Overinflated Sense of Worth k100d Pentax SLR Lens Discussion 11 10-11-2008 10:13 PM
Auto Focus - a bit of common sense garth1948 Pentax DSLR Discussion 2 03-02-2008 06:18 AM



All times are GMT -7. The time now is 08:33 PM. | See also: NikonForums.com, CanonForums.com part of our network of photo forums!
  • Red (Default)
  • Green
  • Gray
  • Dark
  • Dark Yellow
  • Dark Blue
  • Old Red
  • Old Green
  • Old Gray
  • Dial-Up Style
Hello! It's great to see you back on the forum! Have you considered joining the community?
register
Creating a FREE ACCOUNT takes under a minute, removes ads, and lets you post! [Dismiss]
Top