Need your suggestion on the outcome of two test files I've created

[sumptor]

Hey,

I did a little test and was surprised by the results.


Divx file #1:

I captured from a VCR (NTSC) at 320x240 resolution using Huffyuv. I then encoded to divx using Vdub using a 2-pass method at a video bit rate of 1000. The first pass, I kept the resolution at 320x240. However, on the second pass, I added a filter to resize to 640x480. So my final divx file was a 640x480 file encoded at vbr of 1000.


Divx file #2:

I then repeated these steps using the same exact source video but instead kept the resolution at 640x480 from start to finish (ie. captured and performed both encoding passes at 640x480 1000 bit rate with no resizing involved).


I was surprised to see that the first method, with the resizing, looked better visually when compared to the second method. The second file had noticeably more artifacts and had a more "blocky" look to it.

Why is resizing to 640x480 better than staying with 640x480 to begin with?

[wmansir]

just a tip, but you may want to post this in the video capture forum, but anyway..

First of all, Did you deinterlace the full rez capture? Since your capturing NTSC it was interlaced video, so the even lines and the odd lines will be slightly out of sync. This causes a combing effect when there is motion. When you capture at half rez, you only capture half the lines (all the even lines for instance) and so you don't have to worry about interlacing.

If you already know all that and have fixed it (de-interlace/IVTC) then the second reason could just be noise. When you capture at full rez you almost always have to do some noise reduction, unless you have a good digital source, which a VCR is definitly not. By capturing at a low rez and then resizing you are smoothing out the picture, which is a form of noise reduction. If you compared the resized and full rez caps before encoding (in huffy) you should notice that the resized image is smoother, and has less detail. This allows your limited number of encoding bits to create a better looking picture, since there is less information to encode.

[Awake]

@wmansir

Re-read his post, He captured video only once.

Your explanation about capturing only the 'odd' or 'even' fields makes no sense to me, where did you read that? From what I know, both fields will be there regardless of the res, only changes made AFTER capture could change that.

[wmansir]

Reading....

Originally posted by sumptor
Hey,
......
Divx file #1:

I captured from a VCR (NTSC) at 320x240 resolution using Huffyuv......

Divx file #2:

I then repeated these steps using the same exact source video but instead kept the resolution at 640x480 from start to finish (ie. captured and performed both encoding passes at 640x480....


Sounds like he capped twice to me. Once at half-rez and once at full-rez.

To learn more about interlacing read this (http://www.geocities.com/lukesvideo/interlacing.html) from Luke's video capture page (aka Cart, moderator of the vid cap forum). Look at method 3. under the section 'How do I remove Interlacing?" to see what I was talking about.

[Awake]

Hehe, I'm an idiot.....guess I shouldn't post without sleeping for 30 hours. Will be sure to read the link you posted, thanks.