Log in

View Full Version : interlaced source with x264: bob vs. encoding interlaced


me7
4th June 2010, 20:48
I'm going to encode an old TV series that is interlaced and I was wondering what some more experienced users/x264 devs think about the current state of x264's handling of interlaced content.

When I play interlaced TV shows I usually use bob as deinterlacing. This has a major drawback for encoding though, since higher framerate requires higher bitrate. What should I expect when feeding a bobbed PAL source to x264? Will the bitrate rise close to the factor of two? Will x264 use higher quantizers due to the shorter visibility of macroblocks and the bitrate will rise only modestly?

What about encoding it interlaced? Common knowledge says that x264's implementation of interlaced encoding is significantly worse then deinterlaced. Is this still true? Would experienced people or devs expect x264 to handle bobbed 50fps or interlaced content better?

Blue_MiSfit
4th June 2010, 23:10
IMO, encoding interlaced x264 these days is usually more efficient than double rate progressive. Of course, the latter lets you use avisynth deinterlacers, which are insanely good (TGMC etc)! Said deinterlacers can also clean the image somewhat (TGMC), thus improving compressibility.

Long story short, try them both and see what you prefer.

Compare "nicely" deinterlaced double rate progressive at a given bitrate versus interlaced at the same bitrate -using whatever real-time deinterlacing you typically use (aka YADIF in VLC, or your GPU via DXVA or whatever).

Derek

me7
5th June 2010, 19:28
Running tests would miss the goal in this case since I only have this one interlaced source to encode and it happens to be Blackadder, an early 80s TV show with very soft image. There are not many details to preserve between a good and bad encode so a general comparison is out of the question.

I was looking to start a general discussion since there is little talk on the matter. Whenever a newbie asks about interlacing the answer is the same: deinterlace prior to encoding since x264's interlacing is subpar. But x264 has progressed over the past years and there is a theoretical advantage in encoding interlaced and bobbing during playback if x264-interlaced doesn't "cripple" the quality.

I rarely have to deal with interlaced sources. Most of the time it turns out to be telecined but when I have a real interlaced source in front of me I always feel kind of lost and don't know how to deal with it properly.

Blue_MiSfit
6th June 2010, 02:41
Why is a general comparison out of the question with your source? Do the two encodes at identical bitrates and tell us how you like the result :)

~MiSfit

Mug Funky
7th June 2010, 11:33
i've done a test in recent days (2 feature length docos, standards-converted to PAL from NTSC, softish, video sourced).

the results were... very similar.

i think the 50p won out slightly, but it's difficult to tell. in both cases there wasn't a huge amount of movement, so it was tough.

with blackadder, bearing in mind the camera just doesn't move in old UK TV, i'd be inclined to go double-rate and chuck in a couple more bframes.

the blackadder DVDs are atrociously made though - over, over, over filtered, blocky blacks, and mosquitoes over what little movement is in the show. x264 will eat this up no problem. you'll most likely get around 500kbps at crf 20 and a result no worse than the source.

love blackadder. great show.