|
It is currently 02 May 2024, 22:56
|
View unsolved topics | View unanswered posts
|
|
|
|
Author |
Message |
happycube
|
Post subject: Re: (WIP) Laserdisc software image decoder from raw signal Posted: 21 Jul 2013, 17:33 |
Absolute fan |
|
|
Joined: 18 Apr 2012, 18:02 Posts: 1614 Location: United States Has thanked: 71 times Been thanked: 88 times
|
On one level, I know it's quite silly. On another, it's an interesting learning experience for me, and who knows, it might turn out good in the end Even without 'good' results it's a better use of my time than playing video games, etc. Also, once you have a direct scan of the disk and decode it, you get the same kind of RF datastream I'm playing with, just hopefully with much better quality. If someone is in a spot to build a laserdisc scanner and send me complete images of even a few frames of a disk in enough detail, I'd definitely be interested in working with that.
_________________ Happycube Labs: Where the past is being re-made, today. [meep!]
|
|
|
|
|
substance
|
Post subject: Re: (WIP) Laserdisc software image decoder from raw signal Posted: 22 Jul 2013, 06:16 |
Confirmed Padawan |
|
|
Joined: 16 May 2009, 18:05 Posts: 3589 Location: California, USA Has thanked: 28 times Been thanked: 328 times
|
I will capture any LD with my HLD-X0 for you. I have the HLD-X0 routed to JVC HM-DH5U D-VHS deck via S-Video. JVC deck outputs raw digital stream from its firewire port into my laptop. We can try two separate captures. One with all noise reduction turned off. HLD-X0 will apply its TBC and 3D Y/C separation, HMD-DH5U will digitize it. Then we can try one with HLD-X0's built in noise reduction. its highly sophisticated. It has multiple noise reduction adjustments YNR, CNR , VNR, NR. I know YNR is for Luma, CNR is for Chroma noise. I am not sure what VNR and NR does. There is also Sharpness. Unless you pass mid position on these adjustments, video doesn't lose its analog texture. So far it has the best noise reduction I have seen on analog video. It destroyed Faroudja NRS noise reduction, CLD-97's noise reduction, Gennum VXP, TI TVP5160 chip's, Pioneer Elite Kuro noise reduction, ABT2015 noise reduction. I can also try QDEO thru my Oppo BD player but I doubt it can do any better. Only thing comparable to it was Faroudja NRS but it was still better by a margin. It cleaned noise without losing detail, all others blurred the video. You should start off with raw video and try to match or better the noise reduction applied video.
_________________ Coming Soon Derman Labs Anything Of Substance
|
|
|
|
|
Guest
|
Post subject: Re: (WIP) Laserdisc software image decoder from raw signal Posted: 22 Jul 2013, 14:49 |
|
I've pondered building a laserdisc scanner. Optical resolution would be limited by diffraction to a level too low to be useful for precision reading of an LD, so you'd have to use lasers or something like an electron microscope. Someone wrote a very interesting paper about using a CD laser assembly as a profile gauge: http://friedtj.free.fr/CD_eng.pdf If you could do the same thing with a Blu-ray player head, that would be very interesting due to the much smaller spot size. As far as building a scanner - I imagined moving the spinning the disc sideways very slowly under the read head at about 10 revolutions per track, and then capturing the resulting streaming profile data to a large SSD via an FPGA or similar. Being off-center a bit wouldn't matter too much as everything would get rebuilt in software later. Basically you end up with about 1TB of profile data per side, that you can then compress/process. The beauty of this approach is that you can potentially increase the SNR as you should have a 2.5D view of the disc (x, angle,depth) rather than just the 1D view that the laser has, so things like dust particles can be ignored/removed, since you can safely remove anything that doesn't look like a 'legal' laserdisc pit. Also if the pits are a bit deformed you can get a better 'average' size value than the 1D laser would as you would get several passes at different points. You could then fly a virtual LD head over the scrubbed digital disc to get an RF signal, which is where happycube's software comes in 10x slower than real time might be a bit of a problem, so you might need to improve the speed by adding 10 read heads with offset positions so each reads 1/10th of the disc. Of course this is all pie-in-the-sky, the mechanical aspects are pretty daunting, and the resulting signal/image processing software would take man-months of effort, and would probably be a lot slower than real-time to reconstruct the RF signal. Happycube's project seems like a decent middle ground that might actually produce something useful. substance - if you have a moment, and a copy of Video Essentials I'd be very interested to add results for an X0 to my group test here: http://notonbluray.com/blog/comb-tb-tests/ All I'd need is five screenshots from side B: One of the color bars, One of the still zone plate pattern, one of the pluge pattern, one of the Ballerina scene that I reference multiple times in the article (ideally the same frame as the LD-1000/CLD-99 examples), one of the multi-burst pattern near the aspect ratio patterns at the end of side B. Ideally PNGs. Many thanks !
|
|
|
|
|
happycube
|
Post subject: Re: (WIP) Laserdisc software image decoder from raw signal Posted: 22 Jul 2013, 16:40 |
Absolute fan |
|
|
Joined: 18 Apr 2012, 18:02 Posts: 1614 Location: United States Has thanked: 71 times Been thanked: 88 times
|
megapixie - There's a far bigger commercial application for such a scanner - data recovery for scratched/faded CD/DVD-R's. I'm not concerned with real time on LD's, and for CD/DVD data recovery a day would be completely acceptable for most clients.
(Also, I'm surprised nobody in Japan has done what I'm playing with already. LD was more common there, afterall...)
substance - thanks! do you have the Pioneer 8" test disk? I don't have VE yet, but I should soon, so that would be good, and some clips from more average/mediocre disks that the X0 can really work it's magic on. I've heard it mentioned that mediocre players can do well with good disks, but for weaker disks the best players shine.
It's sad the X0 didn't get the LD-S9 treatment in the states, it should have been brought over as the LD-S3.
_________________ Happycube Labs: Where the past is being re-made, today. [meep!]
|
|
|
|
|
Guest
|
Post subject: Re: (WIP) Laserdisc software image decoder from raw signal Posted: 05 Aug 2013, 23:36 |
|
|
|
|
|
publius
|
Post subject: Re: (WIP) Laserdisc software image decoder from raw signal Posted: 06 Aug 2013, 03:12 |
Hardcore fan |
|
|
Joined: 23 Sep 2003, 18:14 Posts: 1391 Location: United States Has thanked: 39 times Been thanked: 21 times
|
megapixie, I'm very interested in what you've done here to extract closed captions with the timing information intact. The video capture software that came with my current capture card will copy just the text to a file, but I'm not going to have the card too much longer anyway. Unfortunately, I'm not terribly familiar with the tools you are using. The way you've put it together, though, it seems as though your method ought to be able to accept a "live" capture stream instead of an already-captured file, & to generate .SSA subtitles instead of .SRT. That would enable me to effectively rip the subtitles off my CC LDs, to use when playing back the Japanese versions. I know that sounds strange, but I have at least three good candidates — the AC3 versions of Wings of Honneamise, Tenchi Muyo, & El-Hazard. How much work do you think that would require to implement?
_________________ MUSE decoder information and user guides LD player connexion guide
|
|
|
|
|
happycube
|
Post subject: Re: (WIP) Laserdisc software image decoder from raw signal Posted: 06 Aug 2013, 08:13 |
Absolute fan |
|
|
Joined: 18 Apr 2012, 18:02 Posts: 1614 Location: United States Has thanked: 71 times Been thanked: 88 times
|
Nice! On my end, I've largely C++11-ized the code, and made some good progress on TBC (it tracks pretty well based off chroma phase as long as it stays close enough) and other filtering stuff. Decoding quality is improved a bit between better captures (using an amplifier and properly insulated cables), better filtering, etc. Looks like I'm getting around 370 lines right now. These are pretty good demonstrations: https://raw.github.com/happycube/ld-decode/master/rd-bw.jpghttps://raw.github.com/happycube/ld-decode/master/ntscbar.jpgA lot of it's disk dependent - the 1985 Star Wars CAV has a nice analog look to it, but VE doesn't look good at all - it has a 2.8-3.5mhz? wave pattern that my filtering hasn't done much good on. I'll capture some other AC3 disks soon and see what's up with that. I'm probably going to do audio decoding as a separate program at least at first. The decoded digital audio signal might provide a good time base, though...
_________________ Happycube Labs: Where the past is being re-made, today. [meep!]
|
|
|
|
|
Guest
|
Post subject: Re: (WIP) Laserdisc software image decoder from raw signal Posted: 06 Aug 2013, 15:05 |
|
publius wrote: megapixie, I'm very interested in what you've done here to extract closed captions with the timing information intact. The video capture software that came with my current capture card will copy just the text to a file, but I'm not going to have the card too much longer anyway. Unfortunately, I'm not terribly familiar with the tools you are using. The way you've put it together, though, it seems as though your method ought to be able to accept a "live" capture stream instead of an already-captured file, & to generate .SSA subtitles instead of .SRT. That would enable me to effectively rip the subtitles off my CC LDs, to use when playing back the Japanese versions. I know that sounds strange, but I have at least three good candidates — the AC3 versions of Wings of Honneamise, Tenchi Muyo, & El-Hazard. How much work do you think that would require to implement? If you can capture to 'live' to an image sequence with your capture card - it should be possible. I could write a generator that watched a directory, and processed whatever files appear, deleting them as it goes. Next on my agenda is outputting in SCC format, which seems to be pretty commonly supported in authoring tools. To be honest, you are probably better off using a third party subtitle format conversion tool that takes SCC as input: My model of the closed-caption state-machine is pretty limited, and I don't have the patience (or the examples) to get things like the coloring, italics and underlining right. Since the SCC format has basically the hex codes of the closed caption stream, any decent standalone converter will do a better job of getting the fiddly details right. I'm not far off having the SCC output done. The problem isn't the work, which is fun, but the time, which is in short supply. Don't forget that the timings generated here are to the frame, so you are going to have to compute some kind of offset between the two discs probably on a per side basis. I their are subtitle tools that can help with that. M.
|
|
|
|
|
Guest
|
Post subject: Re: (WIP) Laserdisc software image decoder from raw signal Posted: 06 Aug 2013, 15:22 |
|
happycube wrote: Nice!
On my end, I've largely C++11-ized the code, and made some good progress on TBC (it tracks pretty well based off chroma phase as long as it stays close enough) and other filtering stuff. Decoding quality is improved a bit between better captures (using an amplifier and properly insulated cables), better filtering, etc. Looks like I'm getting around 370 lines right now. These are pretty good demonstrations:
Looking very nice ! The timebase is looking pretty solid now, still a bit of noise there, but I'm not sure you can do much about that. How does the picture look compared to a frame-cap of the regular output ? I considered suggesting interpolation between the chroma bursts to give a delta term but I think that will probably cause more problems that it solves. I think there are two possible problems with using the audio subcarrier(s) as a timebase reference: * The relative frequency is too low to be useful to detect timebase errors in the higher frequency image carrier * They are 'unlocked' relative to the picture stream (The digital audio EFM carrier particularly) so may drift I would be very happy to be proven wrong. All the best, M.
|
|
|
|
|
happycube
|
Post subject: Re: (WIP) Laserdisc software image decoder from raw signal Posted: 06 Aug 2013, 18:05 |
Absolute fan |
|
|
Joined: 18 Apr 2012, 18:02 Posts: 1614 Location: United States Has thanked: 71 times Been thanked: 88 times
|
Thanks My current TBC works as a two-pass system, the first color decoder provides the phase, and the main loop (for now) does linear filtering - so something like a digital tangential mirror. Right now the value is 1H ahead of time, so there's still more tweaking to do. And it only handles about 1/2 cycle of drift per line. I figured I could extract the EFM bitrate and use it as a correction factor, but it might not actually be viable. Either way the audio decoder will be a separate path, as it's 10X+ oversampled and I only need 44-48K samples/sec, so a regular DFT will provide enough detail. I haven't captured the original picture from this player yet. I think it's also a bit noisy, but not quite as much - this isn't from my best V8000, which a gear on the tray loader broke while I had the tray off to adjust it (which is why they say push the "i'm loaded" button *very* quickly while adjusting stuff) and I haven't gotten around to moving another one onto it yet. The DVD-R capture I have from that V8 has *much* less noise - but ghosting and worse chroma shifting/timing at the edge of the color bars. What my decoding doesn't have, though, is ghosting, which I saw even in the X9 pictures on Magapixie's site. With noise reduction (either from the better V8, another player, or DNR) this could be a uniquely clean image. If someone could look at the comb filtering side of things, that'd be great. I think I'll get a 4fsc/10+bit capture of the SW2 test pattern captured from a VE DVD, to provide a top quality source. I just grabbed an obstensibly working Panasonic BD-35 for $20 at a thrift, so it'd be a good test of that before passing that on to a friend... And I've ordered copies of both those books, thanks Disclord for the info!
_________________ Happycube Labs: Where the past is being re-made, today. [meep!]
|
|
|
|
|
publius
|
Post subject: Re: (WIP) Laserdisc software image decoder from raw signal Posted: 06 Aug 2013, 18:52 |
Hardcore fan |
|
|
Joined: 23 Sep 2003, 18:14 Posts: 1391 Location: United States Has thanked: 39 times Been thanked: 21 times
|
megapixie wrote: If you can capture to 'live' to an image sequence with your capture card - it should be possible. I could write a generator that watched a directory, and processed whatever files appear, deleting them as it goes. Hmm, that might do the trick. Let me see if I can implement that in VLC. megapixie wrote: Don't forget that the timings generated here are to the frame, so you are going to have to compute some kind of offset between the two discs probably on a per side basis. I their are subtitle tools that can help with that. Believe me, that's not the hard part when it comes to timing! As long as the relative timing is right, I can move the start point around to wherever I need it. I already do that with many of the subtitle scripts I have on file, for one reason or another. The only real hassle comes if the side breaks are different, & even that is manageable. megapixie wrote: I think there are two possible problems with using the audio subcarrier(s) as a timebase reference: * The relative frequency is too low to be useful to detect timebase errors in the higher frequency image carrier * They are 'unlocked' relative to the picture stream (The digital audio EFM carrier particularly) so may drift
I would be very happy to be proven wrong. When it comes to time-base correction, it's the relative errors that are more important. So, you can use a pilot tone which has a frequency not locked to the video, as long as it is inherently stable. If the EFM transitions come in with a faster timing, you can bet the chroma subcarrier cycles will too. It might be mentioned that the MUSE Hi-Vision LD format incorporates a pilot tone, although it wasn't ever really used. The idea was to provide for cheap players, since the positive H-sync & lack of colour burst made TBC somewhat trickier than for NTSC, but in the event all or nearly all the players ever built (I'm not sure about the Pioneer HLD-V500) did a digital timebase correction, extracting the necessary timing information from the video signal synchronizing waveforms. The trick for using pilot tone is probably to use it as a reference in a phase-locked loop.
_________________ MUSE decoder information and user guides LD player connexion guide
|
|
|
|
|
Guest
|
Post subject: Re: (WIP) Laserdisc software image decoder from raw signal Posted: 10 Aug 2013, 15:00 |
|
Modest progress. I've now got a python script that can rip visible closed captions to SCC format, which is much more useful, as it turns out: http://notonbluray.com/blog/scc-closed-captions-python/Additionally I've been playing around with software NTSC encoding as a first step to experimenting with comb-filter design. Have got something basic working, but I now need to add various low-pass filters and interpolation. Should have something to show by next weekend. One idea that has been bouncing around my head: software LD+G decoding... Cheers, M
|
|
|
|
|
happycube
|
Post subject: Re: (WIP) Laserdisc software image decoder from raw signal Posted: 17 Sep 2013, 07:42 |
Absolute fan |
|
|
Joined: 18 Apr 2012, 18:02 Posts: 1614 Location: United States Has thanked: 71 times Been thanked: 88 times
|
(belatedly) If you don't have one already, get a Hauppauge HD-PVR. Still working on this on and off, I haven't had much spare bandwidth lately since I moved (and will again next month) What I'm fighting right now is dealing with glitches during sync - I've got 10 second clips, but they suffer pretty badly. The unglitched frames look pretty good, though
_________________ Happycube Labs: Where the past is being re-made, today. [meep!]
|
|
|
|
|
happycube
|
Post subject: Re: (WIP) Laserdisc software image decoder from raw signal Posted: 25 Sep 2013, 07:25 |
Absolute fan |
|
|
Joined: 18 Apr 2012, 18:02 Posts: 1614 Location: United States Has thanked: 71 times Been thanked: 88 times
|
Thanks I'm still plugging away at it, I'll hopefully post a test clip this weekend. (Looks pretty good, aside from the primitive comb filter and poor glitch handling) This morning I went in and finished splitting the TBC from the NTSC frame decoding, for both performance reasons and to get it modular enough for people to play with the comb filtering. The next step is to split the comb filtering bit from the frame decoder. (Since I'm using Linux, splitting it into multiple programs is a very easy way to exploit multi-core too. I'm actually up to about 1/7th real time on video decoding... on an i7.) Then I need to actually comment the code and/or make it clear enough to comment. --- (my lame attempt at a) layman's version: This is a software project that takes the raw signal from the LaserDisc and decodes it into regular video, and then converts that into (hopefully good quality) digital video. Eventually, if someone was able to get a high-quality scan of a Laserdisc's pit pattern, it would be possible to decode laserdisc without any player at all. So in about 10-20 years when LD players will be very rare, it might still be possible to archive them. It's an interesting exercise in learning digital signal processing, etc - and hopefully it might even wind up being useful --- (edit/addition) - it would be theoretically possible to build a AC3-RF* -> HDMI decoder, but that's not a direction I'm planning to go. For one thing, I'm nowhere near real time on an i7 at the moment, and after this project I need to pick ones that are more... profitable?
_________________ Happycube Labs: Where the past is being re-made, today. [meep!]
|
|
|
|
|
|
|
|
|
|
|
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot post attachments in this forum
|
|