[MacDV] Re: Converting 4:3 to 16:9

sb videovideo at mac.com
Mon Oct 10 12:35:45 PDT 2005


Perhaps we mean the same thing.

When you start with a DVNTSC image, 480 lines, each pixel across is on one
line or field. The size of the pixel is a fixed value.
When you scale the image up, (or zoom in) the number of lines or fields
stays the same. Your one original pixel has to be interpolated into 2 or
more pixels, so as the keep one pixel per line.

It's the same when you scale down. You still have the same number of lines
or fields. And, each field or line has only 1 pixel of the same fixed value.

Try this to illustrate:

Look at a frame of video that has motion in FCP's canvas, in Image +
Wireframe. Set it to view at 100%, where you can see the scan lines on the
motion.
Scale up the image using either the Motion tab, or just grabbing a corner
and pulling out. 
Your scan lines don't change. More pixels have been extrapolated, but you
still have the same number of lines.

The amount of jaggies or other artifacting relates to the algorithm used to
create or extrapolate the additional pixels. Better software does a better
job. Scaling up by 33% should be acceptable in FCP, 200% would most likely
look bad.

 regards,

 sb

On 10/10/05 8:29 AM, "Mark M.Florida" <markf at squareblue.com> wrote:

> On Oct 6, 2005, at 6:11 PM, sb wrote:
> 
>> Interlacing doesn't scale, the pixels do. The line size doesn't change.
> 
> Hold on just a sec.  How do you think a frame of digital video is
> interlaced?  Each row of pixels is a different field.  Example:
> 
> Here's your interlaced video signal at 100% scale (1 is field 1, 2 is
> field 2):
> 
> 11111111111111111111111111111
> 22222222222222222222222222222
> 11111111111111111111111111111
> 22222222222222222222222222222
> 11111111111111111111111111111
> 22222222222222222222222222222
> 11111111111111111111111111111
> 22222222222222222222222222222
> 
> If you blow that up 200%, the video will look like this in the same
> viewable area:
> 
> 11111111111111111111111111111
> 11111111111111111111111111111
> 22222222222222222222222222222
> 22222222222222222222222222222
> 11111111111111111111111111111
> 11111111111111111111111111111
> 22222222222222222222222222222
> 22222222222222222222222222222
> 
> So, that's how the interlacing scales.  Interlacing isn't magic -- it's
> just pixels, so if you scale the pixels, the interlacing is scaled as
> well, and you will see the lines prominently, especially in areas of
> motion.

No, this would happen in PS, but not video.
 
> On the other hand, if you *de-interlace* the fields first and THEN
> scale it, it might look better, but de-interlacing involves duplication
> or interpolation of the interlaced lines of pixels, so it's not going
> to look perfect.
> 
> On the other other hand, if you shoot progressive-scan video, there is
> no interlacing and the scaling will look much cleaner.
> 
> Way more than 2 cents.
> 
> - Mark




More information about the MacDV mailing list