question about subpixel motion estimation

[Sh4d3]

Is higher always better?

--level 4.0 --aud --nal-hrd vbr --vbv-bufsize 25000
--vbv-maxrate 25000 --bframes 4 --b-adapt 2 --ref 4
--b-pyramid strict --direct auto --partitions all --weightp 2
--me umh --merange 32 --subme 8 --rc-lookahead 50
--trellis 2 --no-fast-pskip --psy-rd 1.00:0.15 --deblock -1:1

I usually use --subme 8 but I did a few test runs recently using the same settings above, except with --subme 9 and the result looked like crap, especially with dark, less complex scenes where there wasn't a lot of movement. Banding would also become more apparent as this value went up. Is there something I'm missing here?

[Keiyakusha]

Are you sure that after you switched to --subme 9 you encoded file with the same bitrate?

[Sh4d3]

Same bitrate! Actual bitrate used was 12.3Mbit.

[nm]

Could you upload sample clips (source, subme 8, subme 9) that show the issue?

[movmasty]

Are you sure that your encoding app(that you leave unknown) right supports --subme 9 ??

[Keiyakusha]

Are you sure that your encoding app(that you leave unknown) right supports --subme 9 ??

Oo what "support" you talking about? x264 supports up to --subme 11
Or you think that some GUI may be broken to the point that it lists --subme 9 but actually uses different value if you select that? ^__^

[movmasty]

Oo what "support" you talking about? x264 supports up to --subme 11
Or you think that some GUI may be broken to the point that it lists --subme 9 but actually uses different value if you select that? ^__^
Rather than a different value a wrong implementation.

Where r u from Keiyakusha if i may ask?

[Sh4d3]

Could you upload sample clips (source, subme 8, subme 9) that show the issue?

Here's Sample 1 (http://www.mediafire.com/?blil7b8nn5gyup4) and Sample 2 (http://www.mediafire.com/?p3dn256t39izwdc). One note of caution though- unfortunately the previous tests, which exhibited much more noisy artifacts and were much larger in size were deleted. And though the result produced here with --subme 9 is a bit better than in previous tests, I still see very little improvement in quality if any. You be the judge.

[Sh4d3]

Are you sure that your encoding app(that you leave unknown) right supports --subme 9 ??

According to MediaInfo it does!

[Didée]

Well ...

post#1: subme=8 is okay, but subme=9 looks like crap.

post#8: subme=9 shows only little improvement, if any, over subme=8.


Note the difference?

What should anybody tell you? What do you want to hear?