Hallo > >> yuvfps -> yuvmedian -> yuvdenoise -> yuvscaler > >> > >> or > >> > >> yuvmedian -> yuvdenoise -> yuvscaler -> yuvfps > > > >I would use: ... -> yuvfps -> yuvdenoise -> yuvscaler -> ... > > > >You usually do not need yuvmedianfilter any more. > >yuvdenoise checks the differnence between the frames, so you should have > >the needed frame rate before you change the frame rate. The scaler can > >come at laste because it introduces no noise. > Er, shouldn't yuvfps come after yuvdenoise? I have done some tests. I encoded one video with a prograssiv (not interlaced) and one with interlaced frames.
And have encoded the video from PAL to NTSC using that FPS: 30000:1001, 24000:1001, 24:1 using once : fps-change -> denosier -> scaling and the nex try: denoise -> fps-change -> scaling With that Command: lav2yuv file.eli | yuvfps -r xxx | yuvdenoise -f -F | \ yuvscaler -O DVD -n n | mpeg2enc -f 8 -4 1 -2 1 -q 6 -P -I 0 -N -o file.m2v I wondered why yuvscaler had probelms detecting the norm if the framerate as not: 30000:1001. > In the "increasing framerate" case, yuvfps is simply duplicating frames. > If yuvdenoise uses interframe differences in its filtering --- i.e. > frame-to-frame changes are part of its noise model --- then duplication > of frames will make noise look like signal. That might be a probel of my test. I had a rather high quality source, where only litte denoising is needed, that it lookes good. > Conversely, in the "reducing framerate" case, yuvfps drops frames; I imagine > this might enhance the noisiness to yuvdenoise. Here are my encdoded values: for the not interlaced video: Framerate: Order: AVR Bitr. MAX Bitr: 30000:1001 D F S 2593200 4576400 30000:1001 F D S 2611000 4608400 24000:1001 D F S 2304400 4138000 24000:1001 F D S 2305200 4143600 24:1 D F S 2309200 4348000 24:1 F D S 2309600 4349200 orig PAL D 3142800 5401600 for the interlaced video: Framerate: Order: AVR Bitr. MAX Bitr: 30000:1001 D F S 3026000 4949600 24000:1001 D F S 2698200 4298000 24000:1001 F D S 2689200 4308400 24:1 D F S 2664400 4244000 24:1 F D S 2688800 4390400 orig PAL D 3653600 5731200 D=denoise, F=yuvfps, S=scale > Furthermore, if yuvfps is simply duplicating/dropping whole *frames*, it will > probably give questionable results for interlaced streams. (I've never tried > it myself, though.) > Definitely, any deinterlacing (e.g. via yuvdenoise) should come before yuvfps. > > In general, it is always best to do noise removal first --- any good denoising > algorithm is going to be tuned to a particular noise source, and the more > processing you do between the source and the denoising, the more the noise > gets mixed up with your actual signal.... The not interlaced stream was no problem I did not see any problems on the screen. The 30000:1001 PFS worked well in my DVD Player. The interlaced stream looked not that good at the 30000:1001 FPS. If something is moving up the picture it lookes like if you have the wrong interlacing: moving: up, down, up, up up, up, down, up ..... The 24FPS videos did not have that problem (obvious). There I thought sometimes to see it speed up a little bit. So I would suggest to deinterlace before converting it to a other video format. BTW: What do you thinkg about only changeing the header if I would convert from PAL -> NTSC FILM, and streching the audio. The reverse direction is often done for converting film from NTSC -> PAL. Playing back the thing a bit faster, and changeing the audio. What also hit me a little ist that the PAL needs much more bandwith than NTSC. I know that the PAL picture has a larger frame area but I thought the higher frame rate would compensate that. If someone could provide test results the other way round NTSC -> PAL I'd be interrested in the results So I think that I add something to the howto. "How to chnage the frame rate" or something like that. auf hoffentlich bald, Berni the Chaos of Woodquarter Email: [EMAIL PROTECTED] [EMAIL PROTECTED] www: http://www.lysator.liu.se/~gz/bernhard ------------------------------------------------------- This SF.net email is sponsored by: Scholarships for Techies! Can't afford IT training? All 2003 ictp students receive scholarships. Get hands-on training in Microsoft, Cisco, Sun, Linux/UNIX, and more. www.ictp.com/training/sourceforge.asp _______________________________________________ Mjpeg-users mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/mjpeg-users
