Graeme Nattress
RED Problem Solver
I've had the conversation on resolution a few times recently. It's a topic we visited many times in the Red One days and it's just as important today....
The problem is that resolution is multi-factorial and nuanced. If you ignore the nuances you don't achieve understanding. Let's go for some understanding.
First, it's not even resolution we need to think about but MTF. Smaller pixels will lead to a higher MTF even at lower spatial frequencies and thus better contrast handling of the image even when viewed on lower resolution displays.
Aliasing is an issue we want to avoid. Not only does aliasing look bad but it confuses the kind of compression used to broadcast your work making that compression look worse. Nobody wants that. Best way to avoid aliasing is high resolution and a good optical low pass filter.
And all that makes any argument about bayer patterns mostly (or practically) irrelevant as they work more effectively at higher resolutions, and they don't loose you any much more real visible image detail because of that necessary optical low pass filter.
What does low resolution get you? Bigger pixels give you a lower MTF even at low spatial frequencies, and you still need that OLPF because the real world and lenses have much much more detail than your sensor can respond to. Because you are low resolution you will need to / want to sharpen. Electronic sharpening looks bad. That's why there was lots of issues with early HD as it was low enough resolution to need sharpening but not high enough resolution to have real detail without sharpening. And that's one of the reasons diffusion was really popular.
Low resolution means you have little or no downsampling in the creation of that viewing image. Any noise you see (unless you're seeing ghastly plastic skin through over-the-top NR) will be the size of a pixel and even if at a low level will stand out due to its size. On the other hand when you're more massively downsample noise turns to a much more aesthetically pleasing grain.
So yes, the measured resolution of an image will (or should be to avoid aliasing) always less than that of the resolution of the sensor. But this is not something to be used as an argument that resolution is "nonsense" but that resolution is beneficial as downsampling always looks better than upsampling and gives benefits to the image over above and beyond the headline figure of fine detail.
Graeme
The problem is that resolution is multi-factorial and nuanced. If you ignore the nuances you don't achieve understanding. Let's go for some understanding.
First, it's not even resolution we need to think about but MTF. Smaller pixels will lead to a higher MTF even at lower spatial frequencies and thus better contrast handling of the image even when viewed on lower resolution displays.
Aliasing is an issue we want to avoid. Not only does aliasing look bad but it confuses the kind of compression used to broadcast your work making that compression look worse. Nobody wants that. Best way to avoid aliasing is high resolution and a good optical low pass filter.
And all that makes any argument about bayer patterns mostly (or practically) irrelevant as they work more effectively at higher resolutions, and they don't loose you any much more real visible image detail because of that necessary optical low pass filter.
What does low resolution get you? Bigger pixels give you a lower MTF even at low spatial frequencies, and you still need that OLPF because the real world and lenses have much much more detail than your sensor can respond to. Because you are low resolution you will need to / want to sharpen. Electronic sharpening looks bad. That's why there was lots of issues with early HD as it was low enough resolution to need sharpening but not high enough resolution to have real detail without sharpening. And that's one of the reasons diffusion was really popular.
Low resolution means you have little or no downsampling in the creation of that viewing image. Any noise you see (unless you're seeing ghastly plastic skin through over-the-top NR) will be the size of a pixel and even if at a low level will stand out due to its size. On the other hand when you're more massively downsample noise turns to a much more aesthetically pleasing grain.
So yes, the measured resolution of an image will (or should be to avoid aliasing) always less than that of the resolution of the sensor. But this is not something to be used as an argument that resolution is "nonsense" but that resolution is beneficial as downsampling always looks better than upsampling and gives benefits to the image over above and beyond the headline figure of fine detail.
Graeme
