Seeing how the forums been a bit quite lately I though I'd post this bit of info. There's been a fair bit of discussion about de-interlacing and standards conversion.
So I'm about to get some footage run through a high end standards converter and I was loaned the manual to have a read of. Now this nice box made by Leitch does the full bit. To convert between PAL, NTSC and SECAM the fields need to be merged and this is always problematic with motion. This box gets aorund that by using motion compensation. It attempts to derive a set of vectors so it can move the two field back into alignment before they're merged.
Except one critcal adjustment to get this to work right is the noise aperture. Too much noise and it cannot tell what is a deliberate shift in a pixel and what is noise. Fortunately it's smart enough to work out it's out of it's depth and doesn't do motion compensation when it hits that wall.
What does that mean for most of us?
Well I guess just that what you start out with affects the result of many parts of the process. I'm sure one of the reasons the pros get such good results is because of the quality of what they start with. Good cameras with good lenses and decent lighting mean that even after a lot of processing and conversion it still looks OK as all the bits in the chain can work properly. You feed them poor images and they don't just stay 80% OK, you get 80% times 80% which is 64%. Start with 99% and it only goes down to 98%.
The guys who designed the box did include a lot of smarts to get rid of the noise, probably does a better job than most gear but they do warn you hit a limit where the flattening of large areas of similar color becomes very noticeable to the average viewer.
Bob.
So I'm about to get some footage run through a high end standards converter and I was loaned the manual to have a read of. Now this nice box made by Leitch does the full bit. To convert between PAL, NTSC and SECAM the fields need to be merged and this is always problematic with motion. This box gets aorund that by using motion compensation. It attempts to derive a set of vectors so it can move the two field back into alignment before they're merged.
Except one critcal adjustment to get this to work right is the noise aperture. Too much noise and it cannot tell what is a deliberate shift in a pixel and what is noise. Fortunately it's smart enough to work out it's out of it's depth and doesn't do motion compensation when it hits that wall.
What does that mean for most of us?
Well I guess just that what you start out with affects the result of many parts of the process. I'm sure one of the reasons the pros get such good results is because of the quality of what they start with. Good cameras with good lenses and decent lighting mean that even after a lot of processing and conversion it still looks OK as all the bits in the chain can work properly. You feed them poor images and they don't just stay 80% OK, you get 80% times 80% which is 64%. Start with 99% and it only goes down to 98%.
The guys who designed the box did include a lot of smarts to get rid of the noise, probably does a better job than most gear but they do warn you hit a limit where the flattening of large areas of similar color becomes very noticeable to the average viewer.
Bob.