Yes, the fact that resolving power (i.e the system level resolution) is usually measured in one direction at a time, normally horizontally...Or am I missing something here?
No link off the top of my head, but there's lots of info out there. Graeme's posts on these forums are always very informative. The Panavision anti-RED propaganda does offer some insight... It's somewhat accurate in terms of analyzing the shortcomings of a Bayer filter array, but it's also heavily biased marketing FUD in that it doesn't delve into the advantages of a Bayer pattern sensor, nor does it discuss the shortcomings of the sensor tech they have chosen or how the two truly compare. You can Wikipedia or google Bayer Pattern, Bayer filter CMOS, etc.. Anyway, no matter what sort of digital system we are looking at, there will always be a trade-off in actual resolving power or measured resolution in favor of reducing aliasing and other pixellation artifacts, also conditional on the optical path.The part about it recording a 9.44 mp / 4K image that's actually 7.36 mp / 3.3K resolution just 'wooshed' so far over my head I couldn't even see it. :)
Got a link to a Wikipedia article or something?
hehe, cool to see that this topic never goes away ;)
Thanks for the explanations Jeff!
Resolution for digital images is not limited to just the sensor pixel count, it relates to any digital image since the pixels are in fixed locations.
If you take the diaganal line over the digital image you get about x*0.707 and y*0.707 for high MTF limits.
In analog video you can get 1.5 pixels wide, or 1.25 pixels wide spots, but in digital you get 1 or 2 pixels wide and the brightness varies for those two pixels, so video is a little different than digital if the amps had enough bandwidth. And in film the image spots can be anywhere on fine grain film, or at lease average out to being where they should be over a few frames.
So a 2K projector only "resolves" about 1.4K, and 2K on a monitor may "resolve" less then 1280x720 in full RGB true detail at high MTF. You can check the monitor with a black and white checkerboard test pattern for single pixels or blocks of four pixels, you will probably not be able to see the single pixel black and white checkerboard well on your monitor at 1:1 100% size.
Since you are going to lose resolution in projection you get compond losses, so it should be better to shoot 4K and project 2K to stay above 1280x720 rather than shoot 2K and project 2K which would blur the blur more and maybe put you under the 1280x720 value.
Some 2K film recorders only use about 1806x976 so if you do 1806x0.707 you get 1276x690 which is about what people see on prints now, because the film recorders are only "resolving" that much anyway, and you can add losses to that from the film recorder lens, the film printing, and the projector being out of focus.
See these tabels,
It does not matter much so long as you "resolve" more than true 1280x720 which is about what an out of focus 35mm movie projector gives the viewer in the back half of the theatre off a film print.
Other factors such as contrast, highlight and shadow detail, art direction, color purity, etc. will pass through the generations of compression in the release formats more than "resolution". Pixel counts about 8K for movie use to filmout on 35mm would add little in the reduction of aliasing and bayer artifacts, and sensors less than about 2K can show some, so RED ONE having 4K was a good idea and has worked well, as you can see if you go see CHE #1 and #2 in the roadshow prints.
I am glad to see my remarks have sparked such a lively debate, as I thought it would. Reading some of the comments I gather that there is about a 1/4 loss of resolution in the Bayers chip and subsequent processing, or is that an oversimplification? As some of you hinted, I got my info (propaganda) from Panavision.
The Panavision info is a little bit off. But the real problem with it is they don't address the advantages of Bayer sensor implementations, nor do they discuss the shortcomings of the RGB stripe array used by the Genesis.
@ Ken, to assume a 1/4 loss of resolution for a Bayer pattern sensor is indeed an oversimplification. In the case of RED, it's more like an 18~20% loss depending on how it's measured. With most any digital sensor system, it's difficult to put a hard number on effective resolution retained because it varies from one implementation to another. Every step in the image acquisition pipeline plays a role, from the sensor design, to the OLPF, data processing, etc.. One thing that does remain constant however, is that effective resolution is always less that physical resolution within a digital system. Anyone who tells you otherwise is pooping out of their mouth.
|« Previous Thread | Next Thread »|