MPEG Player/resolution

[rjh82]

According to http://www.rockbox.org/twiki/bin/view/Main/PluginMpegplayer#How_To_Encode_Files,

• MPEG-1 and MPEG-2 video is encoded using blocks of 16x16 pixels but the display dimensions can be anything which is a multiple of 2x2 pixels.

I have an iPod photo (220x176); the table recommends a resolution of 220x166 for 4:3 and 220x124 for 16:9.
I notice that the WinFF presets suggest 224x176 for 4:3 and 224x128 for 16:9. 

So,

• What happens when you play a video 224 pixels wide on a 220-pixel-wide display, using MPEG Player?  Does it just throw away four columns?
• If so, does it throw away columns just from the right, or evenly from right and left so that the remainder of the video is centered?
• What are the pros and cons of encoding mpeg2 video in 16x16 blocks versus having fractions left over?  Constraining to 16x16 blocks can introduce a noticeable aspect error at 4:3 (224x176 has an aspect ratio of 1.2727... versus 1.33...).  Whereas, I'm guessing there's a small cost in bitrate when you use a resolution that uses partial 16x16 blocks, such as 220x166.
• Finally, all things considered, is it better for me in general to encode videos 224 or 220 pixels wide?

Thanks!

[Llorean]

Point 1 - Thrown away
Point 2 - From the rightmost edge.
Point 3 - You get better bitrate at the same quality if you use 16x16 blocks.
Point 4 - You really need to test yourself and see if the added bitrate cost (or added loss of quality at the same bitrate) is bad enough to make you pick one way or the other.

[rjh82]

Thanks, that cleared a lot of stuff up.

Point 4 - You really need to test yourself and see if the added bitrate cost (or added loss of quality at the same bitrate) is bad enough to make you pick one way or the other.

I was afraid of that...

[markun]

When I added support for the 2x2 pixel playback in mpegplayer (it used to be 16x16) I noticed that the pixels which will be cut from the right and lower 16x16 blocks (if the resolution is not 16x16), were filled by mencoder by mirroring the pixels from the edge of the image. This avoids some sharp edges (compared to just leaving the pixels black) and should helps to improve the bitrate a bit I think. So maybe the difference isn't very big. I personally just use the 2x2 resolutions.