[Ffmpeg-devel-irc] ffmpeg-devel.log.20121031

burek burek021 at gmail.com
Thu Nov 1 02:05:02 CET 2012


[00:01] <llogan> i think the trac CamelCase autolink behavior may have been confusing to him
[00:03] <michaelni> btw, we have a spam filter in trac, it will detect spam but you guys have to train it
[00:03] <saste> llogan, oh that explains it, sure
[00:03] <michaelni> The training database currently contains 2 spam and 28 ham <--- this is not enough
[00:04] <llogan> ok, i'll feed it some sweet, sweet ham.
[00:04] <saste> llogan, so false alarm, sorry for the noise
[00:05] <michaelni> llogan, "Minimum training required: 25" you need to feed it 23 more spams before it will start working
[00:06] <llogan> i'm slow.
[00:09] <llogan> man, people have some weird-ass usernames. larvalho?
[00:09] <michaelni> also be carefull and dont feed the spam filter wrongly or it will start killing innocents ...
[00:10] <llogan> that's why i'm slow.
[00:16] <burek> does ffmpeg have any free media, like icons, that users can use, when writing about ffmpeg and stuff?
[00:17] <llogan> i'd like herve to re-license the logo. as it is now is too ambiguous.
[00:23] <kierank> that guy appears to be writing about professional sdi cards
[00:25] <llogan> he's legit. just for some reason he decided to make wiki pages to act as link references for his camelcased words since trac automatically assumed they are wiki links.
[00:25] <llogan> no big deal
[03:13] <cone-542> ffmpeg.git 03Michael Niedermayer 07370e39725655: sws: fix handling of big endian alpha
[03:13] <cone-542> ffmpeg.git 03Michael Niedermayer 076c49562d790e: tree-test: fix printf() types
[03:13] <cone-542> ffmpeg.git 03Michael Niedermayer 070e4d34e087e7: pix_fmt_info: add 9/10 bit YUVA formats
[04:45] <cone-542> ffmpeg.git 03Paul B Mahol 07dc239b3b40ad: av_get_audio_frame_duration: add IAC
[05:16] <Compn> 18:11, 27 October 2012 J-b (Talk | contribs) deleted "403.html" ?(Mass deletion of pages added by BoutotNead473)
[07:43] <shmooz> hello!
[07:43] <shmooz> I've been making a video player for android that has lists of channels/streams that users can update, so far I've been using the default VideoView that comes with Android SDK, but it doesn't support enough formats like ffmpeg does
[07:44] <shmooz> so now I have to learn the Android NDK and how to compile ffmpeg 1.0 for it and how to package that with my app
[07:45] <shmooz> anyone here compiled ffmpeg for android? any clues or links would help
[08:22] <cbsrobot> shmooz: maybe this will help you http://fate.ffmpeg.org/report.cgi?time=20121030234947&slot=armv7a-android-gcc-4.4
[08:23] <shmooz> thanks cbsrobot
[08:27] <shmooz> I was looking at some of the older versions which this guy seems to have working as well ..  http://bambuser.com/opensource
[08:33] <cbsrobot> shmooz: well it says: " instructions for building FFmpeg for Android are available for download here:"
[08:33] <cbsrobot> so the instructions should be included
[08:41] <shmooz> cbsrobot: yes, none of those are ffmpeg 1.0, but will be very useful still assuming it will be the same process
[08:45] <shmooz> this sourceforge link also looks promising :)  http://sourceforge.net/projects/ffmpeg4android/
[08:46] <shmooz> that way I wouldn't have to worry about different architectures like arm6 arm7 NEON and so fourth
[08:51] <shmooz> and here's another site to build ffmpeg on android with good instructions ,  http://www.roman10.net/how-to-build-ffmpeg-for-android/
[08:54] <shmooz> I guess I'll try the sourceforge way first because they claim to already have 1.0 supported
[08:56] <shmooz> and the features focus well on making things easy for the developers
[09:38] <ubitux> hum, is it possible to have a bit more insight about draw_slice() callback in lavfi?
[09:38] <ubitux> if i want to process every px i can in this callback, i'm supposed to for (y = y0; y < h; y += slice_dir) ?
[09:39] <ubitux> (y0 being the y parameter of the function, h the height passed as well, along with slice_dir parameter)
[09:40] <ubitux> the draw_slice() callback doxy doesn't contain much info
[10:07] <ubitux> and also, i wonder how well it interacts with code that require a new write buffer
[10:55] <ubitux> ok i did it differently
[11:42] <ubitux> ok, port vf geq done.
[11:44] <ubitux> almost. :)
[12:15] <cone-613> ffmpeg.git 03Stefano Sabatini 073c4aa5062089: lavfi: add field filter
[12:15] <cone-613> ffmpeg.git 03Stefano Sabatini 07652fab5996a3: lavfi/mp: drop wrapped field filter
[12:15] <cone-613> ffmpeg.git 03Stefano Sabatini 071b3fdd973101: lavfi/drawbox: implement color=invert mode
[12:15] <cone-613> ffmpeg.git 03Stefano Sabatini 07003bd88d9961: lavfi/drawbox: reindent after previous commit
[12:21] <cone-613> ffmpeg.git 03Stefano Sabatini 07755038d07d4c: lavfi/drawbox: remove some unnecessary parentheses
[12:46] <ubitux> saste: i ported geq, but i still have a weird issue with the picture not being updated :(
[12:48] <ubitux> i mean, not all the time, it's pretty weird
[12:48] <saste> ubitux, sounds like a permissions issue
[12:49] <ubitux> maybe
[12:49] <ubitux> i think there are two related bugs
[12:50] <ubitux> first one is: using color=..,geq=... it freezes
[12:50] <ubitux> testsrc=...,geq=... it works
[12:50] <ubitux> but i also have a non-changing picture when using testsrc
[12:50] <ubitux> and it's hard to compare with mp=geq since it crashes most of the time :p
[12:51] <saste> ubitux: post the code
[12:52] <ubitux> yup just a sec
[12:53] <ubitux> saste: http://b.pkh.me/0001-lavfi-WIP-vf-geq.patch
[12:53] <saste> ubitux, please on list, now i'm busy
[12:54] <ubitux> i'll do that later
[12:54] <ubitux> i can't really send it right now
[12:54] <ubitux> (i'm not supposed to work on this :p)
[12:55] <saste> ok
[13:04] <ubitux> saste: libavfilter/libmpcodecs/vf_tinterlace.c can't be dropped?
[13:04] <saste> see the patches on the archive
[13:04] <saste> the native version is slightly slower
[13:04] <saste> i tried to optimize it, but there is still some difference
[13:05] <ubitux> ok
[13:05] <ubitux> i just looked at the different telecine & interlacing related filters
[13:05] <ubitux> there are quite a bunch of them
[13:05] <ubitux> i noted detc dint divtc fil filmdint il ilpack ivtc kerndeint mcdeint pullup softpulldown softskip telecine
[13:16] <ubitux> saste: i don't see any particular permission in tinterlace; it's assumed to be read only? otherwise maybe it's already getting automatically an out_buf or something internally?
[13:18] <saste> tinterlace creates and outputs a new buffer
[13:19] <ubitux> yes it does it itself
[13:19] <ubitux> but isn't the internal doing it also in some cases?
[13:19] <ubitux> that's what i'm using in the geq filter btw
[13:20] <ubitux> i'm just working on the internally allocated out_buf
[13:21] <cone-613> ffmpeg.git 03Michael Niedermayer 07976175fbe559: probetest: fix possible use of uninitialized data
[13:29] <saste> ubitux: http://pastebin.com/XGK7Xi8z
[13:29] <saste> my old port of geq
[13:30] <saste> i don't remember why i never posted it
[13:30] <ubitux> !
[13:30] <ubitux> and you never told me you did!
[13:30] <saste> BTW now we should make use of av_opt_set_from_string()
[13:31] <ubitux> doesn't really apply with this filter
[13:31] <saste> it had issues, but can't remember exactly what
[13:31] <ubitux> why is there this copy props & ref/unref thing?
[13:31] <ubitux> you are not referencing any frame manually
[13:31] <ubitux> so why would you unref them?
[13:32] <saste> keep in mind that is old code
[13:32] <saste> so the changes done by anton and nicholas were not considered
[13:32] <ubitux> ok
[13:33] <saste> no, wait, it can't even compile
[13:33] <ubitux> yes
[13:33] <ubitux> just realized that :)
[13:33] <ubitux> the code is still using the old mpi thing
[13:33] <saste> so maybe i never had it working
[13:34] <saste> i got depressed by some problems in mp=geq and gave up
[13:34] <ubitux> anyway, my port is complete, i just have to fix one or two issues
[13:34] <ubitux> yes mp=geq just crashed
[13:34] <ubitux> -d+s
[13:34] <ubitux> it works fine mostly in mplayer though
[13:34] <saste> creation date of the file: 2011-08-05
[13:34] <ubitux> blast from the past!
[13:47] <cone-613> ffmpeg.git 03Mans Rullgard 07d16c4aebba1b: configure: fix print_config() with broke awks
[13:47] <cone-613> ffmpeg.git 03Martin Storsjö 079b50d20cd24c: cngdec: Update the LPC coefficients after averaging the reflection coefficients
[13:47] <cone-613> ffmpeg.git 03Martin Storsjö 07cafefd889b9e: cngdec: Fix the memset size to cover the full array
[13:47] <cone-613> ffmpeg.git 03Martin Storsjö 07036e6c37d31e: cngdec: Make the dbov variable have the right unit
[13:47] <cone-613> ffmpeg.git 03Martin Storsjö 076b68223d315a: cngdec: Allow flushing the decoder
[13:47] <cone-613> ffmpeg.git 03Martin Storsjö 07ab9545a290d1: cng: Reindent some incorrectly indented lines
[13:47] <cone-613> ffmpeg.git 03Diego Biurrun 076860b4081d04: x86: include x86inc.asm in x86util.asm
[13:47] <cone-613> ffmpeg.git 03Michael Niedermayer 073174616f5986: Merge commit '6860b4081d046558c44b1b42f22022ea341a2a73'
[13:53] <saste> durandal_1707, are you OK with the thickness drawbox option?
[13:53] <saste> so I can push and drop mp=rectangle
[13:55] <durandal_1707> yes, i forgot to reply
[13:55] <saste> thanks
[13:56] <relaxed> saste: it would be great if there was a flag to fill the box.
[13:57] <saste> relaxed, t=max
[13:57] <saste> i'll add an example in the docs
[13:58] <relaxed> oh, I don't see it in the man page.
[13:58] <relaxed> thanks then :)
[13:58] <saste> it's a standard (undocumented features) of the option systems
[13:59] <saste> max/min set the max/min accepted value
[13:59] <saste> in the case of thickness, any thickness value big enough will fill the box completely
[14:01] <cone-613> ffmpeg.git 03Diego Biurrun 0704581c8c77ce: x86: yasm: Use complete source path for macro helper %includes
[14:01] <cone-613> ffmpeg.git 03Dave Yeo 07264f12342c3e: x86: Fix assembly with NASM
[14:01] <cone-613> ffmpeg.git 03Michael Niedermayer 079766d9c98515: Merge commit '04581c8c77ce779e4e70684ac45302972766be0f'
[14:06] <cone-613> ffmpeg.git 03Diego Biurrun 07652f5185945c: x86: mmx2 ---> mmxext in comments and messages
[14:06] <cone-613> ffmpeg.git 03Michael Niedermayer 0778ec407d5a7e: Merge commit '652f5185945c8405fc57aed353286858df8d066f'
[14:07] <burek> http://pastebin.com/cwUHmk4M   ...
[14:07] <nevcairiel> you got a smart one there
[14:07] <burek> no comment :)
[14:07] <ubitux> :D
[14:08] <durandal_1707> crazy
[14:10] <ubitux> -Intel Wireless MMX2 Coprocessor: Programmers Reference Manual
[14:10] <ubitux> +Intel Wireless MMX 2 Coprocessor: Programmers Reference Manual
[14:10] <ubitux> huh?
[14:11] <ubitux> wasn't this supposed to be replaced with "MMXEXT"?
[14:12] <nevcairiel> looks like he took the original title from the pdf he is citing there
[14:12] <nevcairiel> which makes one wonder why the hell its named mmxext in the first place =P
[14:12] <nevcairiel> oh wait thats not mmx2/ext
[14:14] <nevcairiel> thats wmmx, which is a thing in ARM world or something
[14:14] <nevcairiel> iwmmx or so
[14:15] <cone-613> ffmpeg.git 03Diego Biurrun 07a65bdceb0606: x86: mmx2 ---> mmxext in variable names
[14:15] <cone-613> ffmpeg.git 03Michael Niedermayer 074b874fc4fc27: Merge commit 'a65bdceb060628881578afb29df4eb222421381f'
[14:25] <cone-613> ffmpeg.git 03Diego Biurrun 07588fafe7f3bd: x86: MMX2 ---> MMXEXT in macro names
[14:25] <cone-613> ffmpeg.git 03Diego Biurrun 07be923ed65901: x86: fmtconvert: port to cpuflags
[14:25] <cone-613> ffmpeg.git 03Michael Niedermayer 0728c0678eb764: Merge commit 'be923ed659016350592acb9b3346f706f8170ac5'
[14:32] <cone-613> ffmpeg.git 03Diego Biurrun 07be2c456e962a: x86: fmtconvert: Refactor cvtps2pi emulation through cpuflags
[14:32] <cone-613> ffmpeg.git 03Dave Yeo 079c167914a1d8: x86: Fix assembly with NASM
[14:32] <cone-613> ffmpeg.git 03Diego Biurrun 075707af8d0b9c: aacdec: Drop some unused function arguments
[14:32] <cone-613> ffmpeg.git 03Michael Niedermayer 0784833b028bf3: Merge remote-tracking branch 'qatar/master'
[14:48] <cone-613> ffmpeg.git 03Stefano Sabatini 0793399e9381e6: lavfi/drawbox: add thickness option
[14:48] <cone-613> ffmpeg.git 03Stefano Sabatini 079d478f38cd26: lavfi/drawbox: add "width" and "height" aliases for "w" and "h" options
[14:48] <cone-613> ffmpeg.git 03Stefano Sabatini 0729a9e4d5c97e: lavfi/mp: drop the mp=rectangle filter
[14:48] <cone-613> ffmpeg.git 03Stefano Sabatini 07f35d5f966f21: lavf/avienc: enable debug log
[14:48] <cone-613> ffmpeg.git 03Stefano Sabatini 07be0b37c649d1: lavf/avienc: fix/extend error message, in case of too large number of skipped frames
[14:53] <cone-613> ffmpeg.git 03Nick Brereton 0709ea482d5dd1: Fix #1827, segfault with XXCH samples when downsampling.
[15:00] <durandal_1707> ffplay -f lavfi -i testsrc=s=640x480,drawbox=c=invert:t=100 <<< looks like inverting only luma is not enough
[15:14] <ubitux> saste: ok don't bother looking at geq, i was doing some crap
[15:14] <ubitux> :)
[15:16] <ubitux> not yet perfect but i fixed one of the issue
[15:19] <saste> ubitux, BTW, you may want to review: doc/filters: add "Notes on filtergraph escaping" section
[15:21] <ubitux> saste: remind me this in 5 hours if i don't do it
[15:22] <ubitux> i'll likely reply in about ~4h
[15:22] <saste> ok, no hurry, i'll leave soon an be online again at late night
[15:24] <durandal_1707> nobody works on porting noise?
[15:25] <saste> durandal_1707, check the archive
[15:25] <saste> I did it some time ago (much time ago), but it was not good enough
[15:26] <saste> maybe I have a local updated copy somewhere
[15:26] <durandal_1707> why not good enough?
[15:26] <saste> slow or something
[15:33] <michaelni> saste, 83938c3d4cb7c2ad4ecaf2dc890b7d366ea85a8b broke opensolaris 
[15:34] <michaelni> (http://fate.ffmpeg.org/report.cgi?time=20121031141755&slot=x86-opensolaris-gcc4.3)
[15:36] <saste> michaelni, why no stderr is shown?
[15:36] <michaelni> saste, no idea, also gmake V=2 does notz work on opensolaris
[15:37] <saste> so scale is crashing, but we don't know why
[15:37] <durandal_1707> it depends on libc implementation?
[15:38] <saste> I suppose so, some way to deal with NULL or something related to strings
[15:38] <saste> ah wait I know (maybe...)
[15:41] <saste> michaelni, I think it is av_log(... "... %s ...", flags_str) with flags_str == NULL
[15:41] <michaelni> simply commiting a fix for that ans waiting is probably the easiest way to test
[15:47] <cone-613> ffmpeg.git 03Stefano Sabatini 071b216421494f: lavfi/scale: avoid to print a NULL value in log message
[16:28] <cone-613> ffmpeg.git 03Michael Niedermayer 0731bf820a2689: av_tempfile: document umask() issue
[16:43] <ubitux> michaelni: "necceary"
[16:43] <ubitux> :p
[17:38] <ubitux> ok i think the second bug is due to a too slow computing
[17:39] <ubitux> though mplayer seems to deal way better than ffplay in that regard
[17:47] <durandal_1707> it sucks that i cant use gdb with clang build stuff
[17:48] <ubitux> anyone has a relatively simple/fast geq equation making use of p() (luma/cb/cr read from the input) ?
[17:48] <ubitux> (for FATE)
[17:48] <durandal_1707> ubitux: you fixed all bugs?
[17:48] <ubitux> yes
[17:49] <ubitux> one that bugs me is that ffplay seems to deal pretty bad with slow eq, but it looks unrelated
[17:49] <ubitux> the rest is fixed
[17:49] <ubitux> i'll send a patch tonight
[17:50] <ubitux> durandal_1707: http://b.pkh.me/0001-lavfi-WIP-vf-geq.patch if you want to test in the meantime
[17:51] <durandal_1707> ubitux: ffplay use separate threads for audio and vide, if video is very slow, audio wins and video frames are not displayed at all, (similar with reading big png files, it reads all of them but do not display any except first one :))) ) 
[17:51] <ubitux> there is no sound in my case
[17:51] <ubitux> just video
[17:51] <ubitux> 'testsrc,geq=p(lt(X\,W/2)*X*(1+sin(N*PI/100))+not(lt(X\,W/2))*(W-X)*(1+sin(N*PI/100))\,lt(Y\,H/2)*Y*(1+sin(N*PI/100))+not(lt(Y\,H/2))*(H-Y)*(1+sin(N*PI/100)))'
[17:51] <ubitux> here is the filter
[17:51] <durandal_1707> nor in png case ....
[17:51] <ubitux> try with ffplay -f lavfi
[17:51] <Compn> yay geq :P
[17:51] <ubitux> (i stole it from a user on #mplayer a few days ago)
[17:52] <durandal_1707> i get segv here with mp=geq
[17:52] <ubitux> yes
[17:52] <ubitux> try with mplayer
[17:52] <ubitux> and/or apply my patch
[17:53] <ubitux> afk for a few minutes, see you
[17:55] <durandal_1707> mplayer is very slow with 720hd
[17:57] <durandal_1707> your equatition is even slower than flip one in manual....
[18:24] <ubitux> :)
[18:25] <durandal_1707> ubitux: couldn't filter use dsp?
[18:25] <ubitux> which one?
[18:25] <ubitux> geq? :/
[18:26] <durandal_1707> yes, but it looks impossible
[18:27] <ubitux> pengvado: are you Loren or i'm mistaken?
[18:35] <thegeek> what does everyone think of the crowfunding proposal?
[18:44] <pengvado> ubitux: yes
[18:45] <ubitux> pengvado: ok, so i have a simple question for you regarding hqdn3d filter
[18:45] <ubitux> any reason to keep the mplayer version?
[18:45] <ubitux> given that you did quite some improvement on the native one
[18:49] <pengvado> I've never tried using libavfilter from mplayer, but if it works then no reason
[18:49] <ubitux> i meant the local copy we have for the mp wrapper
[18:50] <ubitux> libavfilter/libmpcodecs/vf_hqdn3d.c
[18:50] <ubitux> it's in sync with mplayer, but within ffmpeg project it looks quite outdated compared to the native one in libavfilter/vf_hqdn3d.c you improved
[18:50] <ubitux> so i was wondering if there was a reason to keep the old one (like features not ported yet or something)
[18:53] <pengvado> there never was a reason to have libavfilter/libmpcodecs/vf_hqdn3d.c
[18:53] <pengvado> libavfilter/vf_hqdn3d.c existed before libmpcodecs/ was imported whole
[18:53] <ubitux> actually you weren't responsible from the port mmh...
[18:54] <ubitux> right, but libavfilter/vf_hqdn3d.c is based on the mplayer filter
[18:54] <pengvado> first libavfilter/vf_hqdn3d.c was ported from mplayer. at this point it had feature parity.
[18:54] <ubitux> ok perfect then
[18:54] <ubitux> we can just drop it
[18:54] <pengvado> then libavfilter/libmpcodecs/vf_hqdn3d.c was added, and it was already obsolete
[18:54] <ubitux> i see, thanks :)
[19:08] <durandal_1707> ubitux: do not forget to bump micro
[19:11] <ubitux> durandal_1707: ah, good point.
[19:11] <ubitux> not minor?
[19:11] <durandal_1707> not
[19:11] <durandal_1707> see prior removals
[19:11] <ubitux> ok
[19:17] <cone-613> ffmpeg.git 03Michael Niedermayer 073a34ac654174: file: typo
[19:17] <cone-613> ffmpeg.git 03Michael Niedermayer 076992b0bfc2f5: utils: fix rounding of micro seconds
[19:17] <cone-613> ffmpeg.git 03Michael Niedermayer 07adc39b28a526: av_image_copy: check for av_image_get_linesize() failure.
[19:23] <cone-613> ffmpeg.git 03Clément BSsch 07f53112b66582: lavfi: remove hqdn3d libmpcodecs wrapper.
[19:23] <cone-613> ffmpeg.git 03Clément BSsch 075b4cae50c992: LICENSE: update GPL filter list.
[19:23] <cone-613> ffmpeg.git 03Clément BSsch 07fd6707ed0eac: lavfi/hue: add missing GPL dependency.
[19:56] <llogan> what exactly does "updatefirst" do in libavformat/img2enc.c?
[19:58] <burek> can ffmpeg read animated gifs and all the frames within it?
[19:59] <ubitux> burek: there is a patch for this on the ml
[19:59] <ubitux> i don't think it got applied yet
[19:59] <llogan> burek: are you making some new memes for us?
[20:00] <burek> i see :) thanks :) a user is asking on the forum how to extract frames, so i just wanted to check if ffmpeg can do it all-at-once or he needs to extract frames prior to using them with ffmpeg
[20:00] <burek> llogan, not yet :)
[20:00] <burek> still finishing my one-month-work
[20:00] <burek> that looked like it will be a one-week work
[20:00] <llogan> i have some of those too
[20:18] <ubitux> anyone has some fancy geq equation for the examples?
[20:37] <thegeek> ubitux: try rewriting some pixel shaders?
[20:38] <thegeek> http://glsl.heroku.com/
[20:38] <ubitux> :)
[20:38] <ubitux> i'd better take some time to port/write other filters
[20:38] <thegeek> hehe
[20:39] <ubitux> i'd like to finish my convolution filter
[20:40] <ubitux> to allow some basic filtering such as emboss, sharpening, etc
[20:40] <ubitux> and also makes possible for users to specify their own kernel
[20:41] <thegeek> sounds cool
[20:41] <thegeek> would be nice to have for audio as well
[20:41] <thegeek> could do a lot of stuff
[20:41] <thegeek> e.g room correction
[20:41] <ubitux> :)
[20:42] <ubitux> are you proposing yourself for such task?
[20:42] <thegeek> hehehe
[20:42] <ubitux> :)
[20:42] <cone-613> ffmpeg.git 03Paul B Mahol 07010f1ce95443: ffv1: assert that there is at least one slice
[20:42] Action: ubitux wonders when ffmpeg will have motion estimation filters
[20:50] <cone-613> ffmpeg.git 03Clément BSsch 0754a48d2f0452: lavfi: fix references to avfilter_* functions that don't exist anymore.
[21:03] <ubitux> i still wonder if i'm not going to make the filter in the freq domain instead of spatial convolution filters
[21:06] <ubitux> btw, since i'm not familiar with this, when talking about 2d fft, it's basically a fft 1d for rows, and fft 1d for cols, right? 
[21:06] <ubitux> what about 3d fft for images? how does this work?
[21:08] <thegeek> I'm not very good with this stuff myself
[21:09] <thegeek> but yeah, a 2d fft can be implemented with a 1d fft on rows and then 1d on the columns from the first pass
[21:16] <ubitux> any idea about what the 3d fft is doing for images?
[21:17] <thegeek> hm not really
[21:18] <Skyler_> try looking up fft3dfilter or dfttest
[21:18] <ubitux> ok, thanks
[21:19] <ubitux> i remember dfttest being a tritical filter, using fftw
[21:19] <Skyler_> Yeah, both are tricritical
[21:21] <ubitux> i think i'll go with a simple 2d fft to start with, i'll see how to improve things later
[21:21] <ubitux> would be nice to have at least one fft filter in ffmpeg
[21:22] <ubitux> i didn't see any, for video
[21:45] <ubitux> the fft-related ffmpeg api is quite complex
[21:47] <ubitux> i see 3 main categories: fft, rdft, dct; FFT has Complex permute & calc, but also mdct/imdct/½imdct routines, RDFT is for (i)DFT reals, with R’C and C’R in the two directions, and DCT also has 4 different "modes"
[21:47] Action: ubitux is a bit lost
[21:49] <ubitux> i actually have no idea what i'm supposed to use
[21:49] <ubitux> i always played with rdft api, but i wonder if it's the most appropriate for what i'm willing to do
[22:17] <ubitux> heh it might be fun to do something like -vf fft,geq=...,fft=inverse=1
[00:00] --- Thu Nov  1 2012


More information about the Ffmpeg-devel-irc mailing list