...and see that we have, below.
--
Technology has changed and improved. But it cannot recapture detail the AA filter has already disposed of, or that has been lost by a CFA discarding key wavelengths at a partcular location. That information is gone and only hollywood has the algorithms to bring this detail back when the sensor has not received it nor stored it.But the important part here is, the AA filters aren't that exact
either
First of all, as silly as it sounds (for such a simple element of
the imager), the technology has changed over the course of years.
Recent cameras really do behave noticably more differently than
older cameras at and around extinction. The way RAW processing
algorithms handle those frequencies changed a lot too, although it
changed in widely different ways for different companies (which
makes comparisons even more pointless at times).
It is possible for this to occur if the subject matter is right, but not consistant - I never argued some parts were not sharp, just that if the whole thing is not really at that level of sharpness you cannot say it
I don't expect anything other than the behaviour I see today with the SD-10. I'm not looking at miracles, only repetition and consistency of behaviour with an increased pixel count.My bottom point is this - let's not expect interpolated miracles
from SD14's sensor. At least until we can scrutinize some real
world comparisons
Again using the green photosites alone you can tell that 8 is simply too low a number. Even 9 is a bit too low. This is simply an extrapolation of past behaviour where a 3.2x3 Foveon image held somewhat more detail than a 6MP bayer image with 3MP of green photosites. Some arguments exist over how much more detail, and I don't want to go there - I am only saying what I believe the floor on detail equivilence is, here I am totally doing so independant of artifacts.
You can do imaginary comparisons. You can e.g make two different assumptions.
Actually, they aren't. The image is 12.7 MP, the sensor is 12.8 MP. There is no downsampling happening there.
But here's what's wrong with the "extrapolation" and what i've been talking about. It depends on data we've seen from Bayer cameras that are contemporary to the SD10
Just BTW, that's not really a proven trend (AA filters getting stronger, that is). Things aren't that simple - each company has a different policy regarding AA filters and different AA filters get applied to different product ranges (amateur, semi-pro, pro) on top of that. Like i said, i'm pretty much 99% convinced the AA filter in the 5D is much MUCH more weaker than the one 20D's filter was. And 20D seemed to have a bit weaker filter than the 10D.
So the 14 MP JPEG mode is a kind of demo mode? A kind of show case for the technology? A method to make unbelivers understand?
))--
Yes because to me, art is attention to detail and consistency of vision. The differnece to me between a truly great picture and one that is merely nice is paying exact attention to each and every element present in an image, especially color elements in a color image with a mixture of hard lines. As I said the lines coming in and out of detail actually really annoys me almost as much as color issues like that so seeing the antenna as they are in a real print - I would not like it at all.
The pattern is 10 pixels wide - and 27 high. That is close enough to see at what I would consider to be a decent viewing distance. I can tell because it's pretty noticable at 100% on a 19" 1600x1200 monitor, as are issues wtih the antennas.
That does not follow though.
Yes but we all know well the arguments about how B&W resolution charts. These are pretty much the best possibly case for demosiacers and alogrithms that recover detail beacuse every photosite is helping resolve the edges of those lines. However...
From the LPH results I'm willing to acept that, though again I think the real world yields surprizes that make that difference less acacdemic than it would appear from just the charts. That to me is I think the difference bewteen what Phil sees and what Lin sees.
--
More like 6.5MP in the real world. Resolution chart might conveniently favor the tilted matrix, but against even remotely irregular scenes, I think it does little to nothing. I lean toward nothing, because I would prefer the standard bayer 6MP output for it's lack of artifacts.Fuji's interpolation had a very solid reason
Speaking of which, it's interesting to note Fuji S3Pro's
interpolated 12MP had approx. the same resolving power as a
"normal" 9MP bayer, in part thanks to the fact it resolved best in
horizontal and vertical directions (unlike the normally oriented
Bayer which resolves best in diagonal directions).
Sometimes. There are some paintings where you want to see the brushstrokes. And others where it's not so important.
Perhaps. I'd still like you to consider that you might be looking for trouble and finding it whether it really matters or not. (The same thing can be said for excessive attention to X3 aliasing issues.) That said, dcraw is not my favorite converter because of some of these issues. There are other converters that are better at artifact suppression and still have good sharpness.
Think about that again. A percentage increase is independent of the magnitude of the values. 2 vs. 2.34 is a 17% increase. Just as 1 vs. 1.17. Dividing the absolute values by 2 does not change the percent differences.
http://www.dpreview.com/reviews/nikond80/page28.asp
Sure. I would not put too much weight on the absolute numbers, but the amount of improvement over time vs. number of pixels should apply to both color and B&W resolution (unless you are suggesting the algorithms have been tweaked to improve one at the expense of the other...)
Extinction measurements are probably the least accurate and relevant on these charts. They are much more algorithm dependent as you see when you look at the raw results.
Yes, that's why Lin claimed the raw resolution of the D2X was higher than the 5D. But that's just Canon's default in-camera treatment of detail beyond the Nyquist limits.
Well, we also have the user side-by-side raw tests. And they tend to echo Phil's results. But there is also the matter of interpretation. If you find certain imperfections more objectionable than others, you will score images differently.
All JPEG (except where the camera does not support JPEG like the SD19/10.) And all using the "default image parameters".
Not pointless, but certainly of limited usefulness for advanced users. I would venture that still most camera buyers never change the defaults or use raw. However, they've become a smaller and smaller part of his reviews. Most of the time is now spent on the sample scene which is done with both raw and JPEG.
--More like 6.5MP in the real world. Resolution chart mightFuji's interpolation had a very solid reason
Speaking of which, it's interesting to note Fuji S3Pro's
interpolated 12MP had approx. the same resolving power as a
"normal" 9MP bayer, in part thanks to the fact it resolved best in
horizontal and vertical directions (unlike the normally oriented
Bayer which resolves best in diagonal directions).
conveniently favor the tilted matrix, but against even remotely
irregular scenes, I think it does little to nothing. I lean toward
nothing, because I would prefer the standard bayer 6MP output for
it's lack of artifacts.
Check this.
http://www.dpreview.com/reviews/fujifilms3pro/page21.asp
Fujis Claims are even more hot air than some of the Foveon ones I
see here. When put to a fair test, the hot air escapes.
--6.5 sounds a tad to low, though, even accounting for the favourable
matrix orientation. I had an S3 for quite a while back when i still
had a 20D, so i did a few side by side real world comparisons. In
most real world situations you could resize S3's 12 MP output to 8
MP and get slightly more per pixel details than from 20D's "native"
8 MP output.
There were, ofcourse, scenes that looked simply horrible (mostly
irregular textures), but those were a relatively minor percentage.
--More like 6.5MP in the real world. Resolution chart mightFuji's interpolation had a very solid reason
Speaking of which, it's interesting to note Fuji S3Pro's
interpolated 12MP had approx. the same resolving power as a
"normal" 9MP bayer, in part thanks to the fact it resolved best in
horizontal and vertical directions (unlike the normally oriented
Bayer which resolves best in diagonal directions).
conveniently favor the tilted matrix, but against even remotely
irregular scenes, I think it does little to nothing. I lean toward
nothing, because I would prefer the standard bayer 6MP output for
it's lack of artifacts.
Check this.
http://www.dpreview.com/reviews/fujifilms3pro/page21.asp
Fujis Claims are even more hot air than some of the Foveon ones I
see here. When put to a fair test, the hot air escapes.
--------------------------------------------
Ante Vukorepa
My (perpetually) temporary gallery can be found here -
http://www.flickr.com/photos/orcinus/
I don't really buy this 45 degree Fuji stuff.Well, like i said - "in part thanks to the fact it resolved best in
horizontal and vertical directions"
6.5 sounds a tad to low, though, even accounting for the favourable
matrix orientation. I had an S3 for quite a while back when i still
had a 20D, so i did a few side by side real world comparisons. In
most real world situations you could resize S3's 12 MP output to 8
MP and get slightly more per pixel details than from 20D's "native"
8 MP output.
There were, ofcourse, scenes that looked simply horrible (mostly
irregular textures), but those were a relatively minor percentage.
It's the same point I've been trying to make for years: the processing matters a lot!
Which we called the "twisted rope" effect when applied to the SD9/10. None of these sensors has a 100% fill factor.
It's always a compromise between sharpness, detail, and artifact suppression. Not just in the AA filter or even the pixel organization, but also the demosaicing algorithm, upsizing, and sharpening. Nobody gets it perfect, but there are a lot of different approaches that have their own subtle (and not so subtle) differences.
