Video Resizing

Locked
User avatar
jonmartensen
Joined: Sat Aug 31, 2002 11:50 pm
Location: Gimmickville USA
Org Profile

Video Resizing

Post by jonmartensen » Tue Apr 20, 2004 3:00 pm

When I resize a video to a larger resolution, would it be better to use bicubicresize or lonczosresize (with some filters to smooth it out)? I'm increasing the resolution to 1440 x 960 for some of my editing purposes, and will then scale it back down with lanczosresize and filters for the finished product.

I'm wanting to go with bicubic, since it's "soft", but I'd figure I'd ask too.
Image

User avatar
Zarxrax
Joined: Sun Apr 01, 2001 6:37 pm
Contact:
Org Profile

Post by Zarxrax » Tue Apr 20, 2004 3:12 pm

Bicubic isn't soft... just less sharp than lanczos. It doesn't really matter which one you go with, I'd use bicubic since it will compress slightly better.

User avatar
DemonSpawn
Joined: Thu Apr 03, 2003 1:18 pm
Org Profile

Post by DemonSpawn » Tue Apr 20, 2004 3:57 pm

Wouldn't it just be better to leave it at the native resolution? I mean, when you shrink it back down, wouldn't it look worse?

I know it's best to work at the highest possible resolution before exporting as a smaller size, but I've never heard of making something bigger and then making it smaller.
Another pointless post!

trythil
is
Joined: Tue Jul 23, 2002 5:54 am
Status: N͋̀͒̆ͣ͋ͤ̍ͮ͌ͭ̔̊͒ͧ̿
Location: N????????????????
Org Profile

Post by trythil » Tue Apr 20, 2004 4:04 pm

DemonSpawn wrote:Wouldn't it just be better to leave it at the native resolution? I mean, when you shrink it back down, wouldn't it look worse?

I know it's best to work at the highest possible resolution before exporting as a smaller size, but I've never heard of making something bigger and then making it smaller.
SSXSharpen() in mf's SharpTools does just that -- it resizes the video to 4x its size, applies XSharpen, and resizes the video back to its original dimensions.

The idea is that by supersampling, you can give a filter more information, and therefore (in theory) achieve better results.

This obviously doesn't work as well as it does with supersampling as done by modern GPUs -- IIRC, they actually render the scene at 4x its display resolution and then downsample, whereas here we resize up (synthesizing information in the process), filter, and then resize down.

I'm no image processing guru so I can't comment on how well this actually works in practice -- all I know is that I like the results from SSXSharpen() most of the time :)

Locked

Return to “Video & Audio Help”