-
Scaling an image: (technical question)
I know that in FCP, when you move the position of an image (movie, png, etc.) on the y-axis by odd numbers (1, 3, -5, etc.) you get a poorer result in image quality due to shifted field lines. For best results, you should keep y-axis position changes in round, even numbers (0, 2, -4, etc.). But what about scaling an image? Is there a rule of thumb about scaling an image?
I’ve always assumed that with anything digital, it is best to keep things in round numbers, and in factors of 2, 4, or 8 (I assume this because data is binary). But is it the case that when you scale an image in the FCP motion pane, that it is better to use a factor or 4? Would there be a quality difference in an image that is 87.5% (reduced by 1/8) versus 88% or 90%? Or 112.5% (increased by 1/8) versus 112 (increased by 3/25) or 110% (increased by 1/10)? I’m just wondering what happens to the pixels when they’re being displaced by odd-fractions that don’t mesh with binary factors of 4 & 8.
This is a fairly technical question, but I am curious if anyone knows whether these little math differences matter when it comes to quality… and what’s the best rule of thumb?
-
3 Replies
-
I have no idea why there is such inferior scaling in final cut pro.
premiere, vegas and other NLE’s all have bicubic resizing that works
very well.it’s a weakness for sure.
-
[Jason Diebler] “I know that in FCP, when you move the position of an image (movie, png, etc.) on the y-axis by odd numbers (1, 3, -5, etc.) you get a poorer result in image quality due to shifted field lines. For best results, you should keep y-axis position changes in round, even numbers (0, 2, -4, etc.). But what about scaling an image? Is there a rule of thumb about scaling an image?
I’ve always assumed that with anything digital, it is best to keep things in round numbers, and in factors of 2, 4, or 8 (I assume this because data is binary). But is it the case that when you scale an image in the FCP motion pane, that it is better to use a factor or 4? Would there be a quality difference in an image that is 87.5% (reduced by 1/8) versus 88% or 90%? Or 112.5% (increased by 1/8) versus 112 (increased by 3/25) or 110% (increased by 1/10)? I’m just wondering what happens to the pixels when they’re being displaced by odd-fractions that don’t mesh with binary factors of 4 & 8.
This is a fairly technical question, but I am curious if anyone knows whether these little math differences matter when it comes to quality… and what’s the best rule of thumb?”
I just spent 20 minutes Googling this. It seems nobody knows what scaling algorithm FCP 6 uses. I have one post here on the Cow saying FCP 3 used bicubic while AE back then used bilinear (meaning FCP was better than AE in 2003 at least). But maybe Apple changed it to bilinear make RT Extreme work in FCP 4. Who knows. FCP 6 has three different quality settings for scaling. Is “best” bicubic? I’d love it if someone knew definitively.
I know Compressor uses Lanczos which is supposed to be the absolute best for video. Better than bicubic.
As for all the math and field questions, I think FCP 6 is smarter about that stuff now. But a progressive workflow always eliminates those concerns.
Sean
-
[Sean ONeil] “FCP 6 has three different quality settings for scaling. Is “best” bicubic? I’d love it if someone knew definitively.”
if it was bicubic it would scale a lot better.
all 3 settings act like bi-linear to my eyes.
dang, if apple just implemented Lanczos that would be awesome!
Reply to this Discussion! Login or Sign Up