edges, esp sharp ones look yuck compared to divx?

andy

C:\Program Files\x264\x264.exe --pass 3 --bitrate 1357 --stats quot;.statsquot; --ref 16 --mixed-refs --bframes 3 --b-pyramid --b-rdo --weightb --filter -1,-1 --subme 6 --analyse all  --8x8dct --me esa --merange 32 --cqmfile quot;C:\Program Files\x264\eqm_avc_hr.cfgquot; --progress --no-psnr took about 2 weeks to encode (3 passes) and the quality is no better than a 2 day encode.

The quot;problemquot; is edges and high low contrast bits look like they've been dropped into a lower resolution (circular bits look jagged).

I did a 3 pass divx encode with max quality same bitrate, and it looks heaps better.

Am I wrong to be expecting x264 to shine?  (I did try very hard to RTFM with google and here as best I could)

C:\Program Files\x264\x264.exe --pass 3 --bitrate 1357 --stats quot;.statsquot; --ref 5 --mixed-refs --bframes 2 --b-pyramid --b-rdo --weightb --bime --filter -3,-3 --subme 6 --analyse all --8x8dct --me umh --merange 16 --trellis 1 --no-fast-pskip --cqmfile quot;C:\Program Files\x264\eqm_avc_hr.cfgquot; --progress --no-psnr

That should be much faster and should clear up your issues. I also recommend you go with 2 passes as opposed to 3 if your computer is that slow. The quality will not suffer much.

In addition, never use quot;--me esaquot; it is an exercise in futility.
No need to use more that 5 references.
Not need for merange to be more than 16.

Speed is not a real worry for me......just as long as the end result is of good quality.

2 weeks, that's a waste of time

What's the resolution of your source??

C:\Program Files\x264\x264.exe --pass 2 --bitrate 1357 --stats quot;.statsquot; --ref 3 --mixed-refs --bframes 3 --direct spatial --b-pyramid --b-rdo --bime --weightb --filter X,X --subme 6 --analyse all --8x8dct --no-fast-pskip --trellis 2 --progress --no-psnr

How about this, will be much faster

Oh, forgot about the decoder, use ffdshow and turn on postprocessing and how it corrects the jagged circle you mentioned, coz divx has postprocessing on, different postprocessing settings will also make the picture look different

Postprocessing MUST be turned OFF!
h.264 has already inloop deblocking and doesnt need postprocessing.

Jagged edges might be an overlay problem. Software resizer is simply nearest neighbour. What decoder and player are you using?

you do know that --pass 3 doesnt mean quot;third passquot;?
if you used --pass 2 before --pass 3 you only did a two pass encode...
but it can be questioned anyways whether 3 passes is worth the time

(also 16 reference frames is a not needed overkill imho)

Postprocessing MUST be turned OFF!
h.264 has already inloop deblocking and doesnt need postprocessing.

thanks, another new things learned today

that's unless you deactivate inloop deblocking...

And a ME_range higher than the defaut will most likely decrease slight the quality...

Use the Sharkthoot's profiles or see this tread for better choice the options in relation to speed.

Oline 61,

I didn't get any tangible improvement with those settings.

kotrtim,

I simply do not care for the time aspect.  Quality is the issue here.

And I did try Shartooth's profiles beforehand and all of these settings yield yucky edges.

Please at least give me a firm yes or no on this:  Should x264 @ 1357 bits be better quality than divx 6 or not?

thnx!-cc

It should be much better quality. There is probably a problem in your decoder. What decoder are you using? Is there any post processing done?


Originally Posted by createcomsAnd I did try Shartooth's profiles beforehand and all of these settings yield yucky edges.

Can you post a screenshot that shows the problem?  Your vague description of quot;yuckyquot; isn't much to go on.

wouldn't 16 reference frames make sense with the --mixed-refs option?


Originally Posted by createcomsThe quot;problemquot; is edges and high low contrast bits look like they've been dropped into a lower resolution (circular bits look jagged).

Looks it similar to this?Bye

wouldn't 16 reference frames make sense with the --mixed-refs option?

Forget that. The gain is too small, and the time too big. 5 or even less (3 or 4) references shoud be suficient for most videos.

divx6:x264:
Looking at the images I would say they didn't use the same source. The x264 source seems to be processed by some deinterlacer ( a bad one btw).

--pass 2 --bitrate 736 --stats quot;.statsquot; --ref 3 --mixed-refs --bframes 3 --direct spatial --b-pyramid --b-rdo --bime --weightb --filter -4,-3 --subme 6 --analyse all --8x8dct --no-fast-pskip --trellis 2 --progress --no-psnr

dgindex - forced film
this is a fulll length 1 hour 44 min taxi, I'm not extracting a small portion and encode..

720x368

haha... so much faster than yours, even better...more details
Everything about the encode between those two is IDENTICAL except for the codec used......

kotrtim, if your post is supposed to be helpful I'm sorry but I've failed to appreciate it?