Putting quality emphasis on centre of video

[PlazzTT]

I have made backups of a lot of movies which are in 2.35:1 aspect ratio.

I find that this is too wide for me to enjoy on my 17" CRT monitor, so I usually zoom in on the centre of the video with Media Player Classic, so that the aspect ratio of what I'm viewing is closer to 1.77:1 (16:9).

What bugs me when I do this though, is that bits are wasted on encoding the extreme left and right of the frame, which I don't see because I've zoomed in on the centre.

My question is, would there be any point in a feature that wouldn't "waste" as much bits on these parts of the movie. ie- focus on the middle 75% of the frame? (Something similar to the "credits" feature, only one that would work on specific areas of the frame, increasing the quantiser in areas of the frame which are not as important)

Or is this idea just crazy? I suppose the transition line between the centre of the frame and the extreme left and right parts wouldn't look too good.

Any thoughts?

[Sharktooth]

What about cropping before encoding?

[Brother John]

Afaik no codec can do it. As you don’t care about the extreme lefts/rights cropping seems a reasonable solution.

Nonetheless, the idea is interesting. If you watch a movie you certainly concentrate your attention on the most important part of the picture. So lower quality (within certain limits) in the not so important sections should go unnoticed.
The problem would be finding the important sections. Most of the time the important part of the picture is indeed its center, but that’s not always true. How should that complete idiot of a computer decide what’s important an what’s not? After all (decoded) video frames are nothing more than a set of colored pixels. They only gain their meaning in our brain.
A computer would need something similar to human intelligence for finding the important picture parts automatically. We got a fantastic team of programmers working on XviD but "XviD AI edition" definitely seems a long long way to go. Compared to the possible savings in file size/higher quality in important sections it would hardly be worth the effort.

[Teegedeck]

Thought about that, too but it's not really such a cool idea once you think it through - because most of the time, you don't actually concentrate on the middle of the screen but on that section where the action is happening or where there is an interesting detail, whichever corner of the screen that may be. I say, crop your picture.

There are more interesting ideas that still are exploitable; like 'motion masking'. All in good time, I hope.

[celtic_druid]

Guess you could filter the outside 16/32x??? or whatever sections seperatly, making them more compressable/less detailed.

[scharfis_brain]

I've read about MPEG-2 Encoders (Hardware), that are able to do more quantisation on the borders than on the centre of the video.

this is mostly used of DVB.

(if you look closer to some DVB-clips, you'll notice, that they are much more blocky on the borders.)

so, I think, this should be possible for software encoders, too. Even for MPEG-4 & derviates

[Sharktooth]

It IS possible (something like adaptive quantization). But it's not applicable with all movies/sources/whatever.

[RadicalEd]

Been discussed previously here (http://forum.doom9.org/showthread.php?s=&threadid=73397), sysKin says it'll be possible when the HVS plugin system is implemented.