Re: Compression and PSNR
- From: Thomas Richter <thor@xxxxxxxxxxxxxxxxxxxxxxxxxx>
- Date: 26 Apr 2005 08:16:17 GMT
Hi,
> I have my compression scheme and I want to compare its performance with
> MPEG1. Now the problem is that my scheme produces compression frames
> with certain size (my_size) and provides some PSNR value (my_PSNR) for
> certain settings. Similarly MPEG gives its own compression
> size(mpeg_size) and PSNR (mpeg_psnr) for same settings. Now is it all
> right to compare the ratios my_size/my_psnr and mpeg_size/mpeg_psnr and
> conclude that the scheme with the smaller ratio is better?
It is at least a starter. I would measure PSNR against bpp (bits per
pixel, i.e. the average number of bits required to encode a frame).
Note however that PSNR does not exactly correlate very well with
visual performance. It is very well agreed upon as quality measure, and
it is very well agreed that it does not work too well. I would say that
visual tests (i.e. humans watching your streams and compare them with
MPEG1) should have the last word.
> Is there any other method to compare the results incase this is not the
> correct approach?
Compress to the same bpp value, use test persons and ask them to evaluate
the stream. Continue the experiment with streams of various types.
So long,
Thomas
.
- Follow-Ups:
- Re: Compression and PSNR
- From: roshu
- Re: Compression and PSNR
- References:
- Compression and PSNR
- From: roshu
- Compression and PSNR
- Prev by Date: Re: [q]image combination?
- Next by Date: Re: Compression and PSNR
- Previous by thread: Compression and PSNR
- Next by thread: Re: Compression and PSNR
- Index(es):
Relevant Pages
|