resizing quality loss?

[AlanHK]

I need to reencode some videos, to IVTC for one thing.

I'm using Avisynth and MeGui to make x264 for MKVs.

Some of them have odd sizes, I would like to resize them all uniformly. Is there any loss of quality in doing that, or is the original resolution going to give me the best quality?

I know of course that you can't create high quality high resolution from a low res source, but say going from 704x476 to 640x480? (These are 4:3 DAR.)

[sven_x]

It depends on the optical image resolution of the source, which can be quite lower than the pixel resolution.
There may be ways to identify optical resolution by means of fourier transformation.
But I prefer a much simpler approach.

1. Just take a screenshot of a sharp scene of your video.
2. Downscale this screenshot with an image processor (Irfanview, photoshop, Gimp) to the desired resolution (640x480).
3. Upscale the downscaled version back to the original resolution and save the image.
4. Than just switch between the two screenshot versions. In many cases you will see no difference.

[lisztfr9]

Well, in the Gimp for example, to compare two images, set the upper layer to "grain extract", (layer dialogue, mode), this will highlight any differences in luma.

For resizing, consider bilinear vs lancosresize, the later is supposedly sharpen a bit. All being good quality.

L

[AlanHK]

It depends on the optical image resolution of the source, which can be quite lower than the pixel resolution.

Thanks, I suspected something like that, though I didn't know the terms.

Anyway, I did a test and there doesn't seem to be much if any quality difference between different output sizes.

[creaothceann]

You can just specify the aspect ratio in the MKV properties (using MKVmerge GUI) and the player will resize to that, unless the video stream itself has embedded pixel ratios.

[AlanHK]

You can just specify the aspect ratio in the MKV properties (using MKVmerge GUI) and the player will resize to that, unless the video stream itself has embedded pixel ratios.

I know that. Anyway, my TV AR settings are manual, regardless of what the file DAR is.
Since I was reencoding anyway I thought I might as well make the AR more rational.