11-26-2016, 11:08 AM
|
|
Free Member
|
|
Join Date: Nov 2016
Posts: 2
Thanked 0 Times in 0 Posts
|
|
Dear forum staff and members,
Over the years I already learnt so much from this site that this alone was reason enough to finally become premium member.
I am more of an IT guy but I often go through projects in my free time touching areas of video including analogue video.
In the past I now and then captured VHS through diffenrent devices like an old PCI TV card and such, but I never really achieved good quality and the last time is many years ago.
Inspired by the question of a collegue und another one of my parents, I dove into "prosumer"-VHS capturing and now made it my project to set up a workflow to capture the tapes of my family and friends.
As this is sometimes an advice: Yes probably it would be economically more wise to simply send them in to some professional service, but one of my main motivations is to do it my self as good as possible. This is the way I usually learn the most in other areas too.
Through the forums as I said I already learnt a LOT. For example what to watch out for with the playing device and such. In this thread I hope to get some of my questions answered about the process from the capturing device to some intermediate format for further processing.
So as it's 2016, I live in europe and use linux I feel like I didn't have much choice about the capture device/video grabber.
I tried some Terratec devices, one of AVerMedia and now chose a Hauppauge USBlive-2, which went well with my operating system and seems to give slightly better results than the other ones.
Now one of the most backed facts is that the video should stay interlaced as long as possible. E.g. my parents need DVDs as results as they don't use a SmartTV or a computer to watch videos. (My collegue on the other hand would be fine with some h264-MPEG-4-File.)
I hope some of you already got what my problem might be:
Those USB-grabbers don't send 50 fields per second to my operating system but instead 25 interleaved frames. (With comb...)
Is the information advantage of interlacing already gone at this point?
How would I capture 50 fields instead? I have USB3 and think there are 50fps-grabbers out there, but would that be of any help?
(I guess the result would be like bobbed deinterlacing, but as they are built for HD material I wonder if they cope well with VHS. Also will I get software to convert this back for interlaced DVDs?)
Then I have a question about the intermediate format:
I know Huffyuv uns UTvideo and up until now I used them. If it only means a very slight loss of quality I would probably like to use another intermediate format such as h264 with crf between 8 and 18. Now the most matching color format libx264 supports would be YV12, while I think vhs should be captured in YUY2. Is there a difference in the result? I can set the grabber to each of those. Or can I use another close-to-lossless encoding than this? I think VP8/9 supports lossless and probably close-to-lossless, but I'm not so sure about interlacing.
Right after grabbing there's another question:
For all the processing most of you seem to recommend avisynth while on linux it's far easier to use ffmpeg/lavc.
I think I won't buy a full field tbc, but I own a dvd recorder wich get's rid of some of those problems when I take the signal through it. I hope to get close to a good result with software filters after that.
Obviously at least for the DVD conversion I need decent denoising and every remaining quality issues that I can get rid of through software filters would be great.
From what I've tried until now about denoising filters of ffmpeg hqdn3d is fine while nlmeans unfortunately takes far too much runtime for the number of tapes I want to transfer. So if I sticked with ffmpeg I would use hqdn3d most of the time and nlmeans for some very precious tapes.
From some threads in the forum I get the impression that avisynth is far superior and if it could even make terrible parts of tapes look fine again and could denoise captured video so many details would remain after MPEG-2 compression for DVD. Is there that much of a difference? I think I could include it in my workflow through wine, but it would take a little effort, so I'ld like to know if it's worth the effort.
Then as I am a premium member now maybe I am allowed to ask a stupid feeling question. Most of my work already might have been not important because of the interlaced/interleaved usb grabber thing, but there's something I simply don't understand:
(All conversions came from a utvideo-file as it was captured with YUY2 from the USBLive-2) - Conversion to MPEG2 with settings "il:me,il:dct, top field first" resulted in a smooth picture, Media Info says "interlaced, top field first"
- Conversion to MPEG2 with settings "il:me,il:dct, bottom field first" resulted in a jumping picture, Media Info says "interlaced, bottom field first"
- Conversion to MPEG2 with settings "deinterlacing, il:me,il:dct, top field first" resulted in a smooth picture, MediaInfo says "interlaced, bottom field first"
So I guess when I'm capturing those PAL tapes they're top field first. Are all PAL tapes this way? Or does it depend on the usb grabber? Or ist it completely random for every capture process?
Next I wanted to ask, why the last video has "interlaced, bottom field first" information in Media Info. Now I guess it's because I deinterlace and then interlace again, still it doesn't make much sense. If this was a stupid thing the test that doesn't bring us further maybe we should ignore this file
For the captured UTVideo-File it only says 25fps and nothing about interlacing. So do I currently capture non-interlaced/interleaved video and then reinterlace id? Is this a good idea?
Kind regards
fakir
|
Someday, 12:01 PM
|
|
Ads / Sponsors
|
|
Join Date: ∞
Posts: 42
Thanks: ∞
Thanked 42 Times in 42 Posts
|
|
|
11-26-2016, 01:01 PM
|
|
Premium Member
|
|
Join Date: Aug 2009
Location: N. Carolina and NY, USA
Posts: 3,648
Thanked 1,322 Times in 991 Posts
|
|
Welcome to digitalfaq. Just to clear up a few matters at the outset:
Interlaced PAL video is 25 frames per second but plays back at 50 fields per second. VHS is usually top field first, DV is usually bottom field first. Capture VHS at 4:2:2 YUY2, not YV12. Huffyuv and UT Codec are lossless compressors -- there is no quality loss. You can decompress and recompress lossless codecs forever without quality loss. h264 is a lossy encode with quality loss at the outset, and each recompression involves further quality loss. The loss is permanent and cannot be restored. Capturing to final delivery lossy formats (h264, MPEG, etc.) is capturing to a format not designed for modification. Filtering and denoising lossy final delivery encodes cannot be done without without further re-encoding and degradation. I don't know where you're getting your information about lossless codecs and interlacing, but it's incorrect and didn't come from this forum.
Combing is obvious in editors and playback devices that don't deinterlace. Meanwhile I can't imagine how you can view anything that comes over your cable TV if you have to mount those interlaced broadcasts in your computer and deinterlace them before viewing. Do you see interlaced combing during TV shows or watching DVD movies? If so, something's wrong. Don't you let the playback systems deinterlace, as they were designed to do? Capture software cannot deinterlace well at all, period.
Nothing beats avisynth for denoising and repair. ffmpeg is an inferior and limited way of doing it.
I don't know wherew you heard that a bob operation is designed "for HD". Any interleaved video can take a simple bob, which is an inferior form of deinterlace that simply separates interleaved fields and does a crude resize of half-height fields to full frames. Smart bobbers like yadif, nnedi3, and QTGMC do it differently. Many playback devices designed for interlaced and telecined video will treat it as interlaced anyway. You can deinterlace to 50 frames per second playback, but 50 fps is not valid for DVD or for many BluRay formats. The only HD BluRay format that can accept 50fps progressive is 1280x720, other BluRay formats from SD to HD are either interlaced, telecined, or run at 23.976/24fps film speeds. Standard definition PAL BluRay is 25fps interlaced. If you are deinterlacing during capture by using field-blend methods, you're ruining your videos.
Last edited by sanlyn; 11-26-2016 at 01:45 PM.
|
The following users thank sanlyn for this useful post:
fakir (11-27-2016)
|
11-27-2016, 02:18 AM
|
|
Free Member
|
|
Join Date: Nov 2016
Posts: 2
Thanked 0 Times in 0 Posts
|
|
Hello sanlyn,
thanks for your answer. I think I got too excited about finally posting here so I wrote too much, maybe I should edit and shorten it, what do you think?
Seems from my silent reading all the time I got some things wrong, thank you for clarifying!
Quote:
Interlaced PAL video is 25 frames per second but plays back at 50 fields per second
|
I think you closed a big gap here! So if this usb grabber sends 25 frames per second, the precious information provided through the interlacing is not lost as long as I keep in mind that it's tff and interlaced?
Quote:
Capture VHS at 4:2:2 YUY2, not YV12.
|
Okay, so YV12 would be a loss of information? Then we won't need to discuss h264 anymore. (h264 has lossless setting crf=0 which from my understanding should work about like MJPEG, but as it only supports YV12 it's not lossless for me. I wondered if YV12 would only be another way of storing the same information, thus my question.)
Quote:
(Paragraph about deinterlacing.)
|
I understand that and I try not to deinterlace. I think this route only came to me when I thought that the device would already lose the interlacing information, you clarified this in the first quote, thanks for that. I don't have cable TV for a while now and the tapes are not broadcasted shows but personal videos e.g. of a birthday or a marriage, all in PAL and all shot with consumer grade cameras, which I also understood doesn't make things easier. (Some are Hi8/Video8, but I wanted to focus on VHS first.)
With all that information we can also ignore my thoughts on 50fps grabbing, as it wouldn't bring the benefit I thought of.
If you care I try to explain it better than in my first post:
What I thought and might be wrong: - I think I read in some old thread - probably not here but at gleitz or doom9 - that people tried to capture with half frame height on 50fps, so they get every field as a frame.
- I thought that bobbing would be a technique were the 50 fields would be saved as 50 frames with gaps between the lines.
So I thought if my grabber doesn't "capture 50 fields per second but 25 frames and thus loses information" ( which again I now know is wrong thanks to your answer), I should get a grabber which is capable of capturing 50 frames per seconds to capture every field as half a frame.
Those 50fps grabbers are what I thought would be meant for HD video.
Then of course I would need software which is able to convert the result to 25fps real interlaced material. So I thought this process would be like "de-bobbing".
Again, now that I now that I my most important assumption is wrong, this whole thought is probably worthless.
So what would be the best way to go on? I guess I need to try to set up avisynth through wine and come back with samples, for recommendations for the filter chain?
About the forums, not my project:
All the things I learned are all the things I don't need to ask anymore, which is a lot
Oh my god there's a typo in the threads' name and also the topic shifted a little, can I edit that somehow?
Regards,
fakir
Last edited by fakir; 11-27-2016 at 02:39 AM.
|
11-27-2016, 03:20 AM
|
|
Premium Member
|
|
Join Date: Aug 2009
Location: N. Carolina and NY, USA
Posts: 3,648
Thanked 1,322 Times in 991 Posts
|
|
Quote:
Originally Posted by fakir
there's a typo in the threads' name and also the topic shifted a little, can I edit that somehow?
|
I see typos often -- my posts have some very embarrassing ones. Unfortunately posts can't be revised after an hour. But you can email a moderator and request a correction.
I can advise about Avisynth and VirtualDub as far as I'm able (no member here knows everything, you see, but public forums are open to all who can contribute). I can't advise about Linux, which I don't use. I have software that would never run in a Linux OS and I still use XP for most restoration work.
We recommend VHS capture and cleanup to be done the way the pros do it. Few of us can afford professional software or hardware, but we can adopt their basic principles. For VHS capture, restoration, repair, and encoding, pro methods are based on capture and work with lossless media and lossless codecs such as huffyuv, Lagarith, or UT Video Codec. Video source that is interlaced is maintained as interlaced to preserve initial motion quality and to avoid artifacts should deinterlacing be required later. Capture is at YUY2 because that colorspace is most similar to the way VHS is played as YPbPr and YUY2 does capture more chroma information. True, if your final destination is any standard encoding such as h264 or MPEG the result will be YV12, but I would prefer to leave colorspace conversion to apps like Avisynth which do it with more precision.
The capture and restoration forum threads where samples and actual working scripts are posted would be the best areas for reading about the travails of capture and how they're repaired or managed. Remember, too, that those threads usually address unusual problems and are not generally indicative of "normal" procedure or requirements. Unfortunately when things go well or routine with VHS work you usually don't hear about it. It's front page news only when something goes amiss, LOL! However, many members do submit original capture samples for evaluation about their equipment and procedures. Most lossless capture is done with the VirtualDub capture utility, although we find other software occasionally mentioned.
|
12-03-2016, 06:28 PM
|
|
Free Member
|
|
Join Date: Nov 2015
Posts: 54
Thanked 10 Times in 8 Posts
|
|
I use Handbrake for making deinterlaced MP4 ( from interlaced MPEG2 ) because many computers cannot show MPG or only shows it interlaced. However MPEG2 is sometimes best for playing on tv via USB, especially old slower models.
https://handbrake.fr/
Free program also for Linux
There is a lot of settings in Handbrake, after some experimenting i use this
Picture: Anamorphic = Strict, Cropping = Custom, left=4, right=10, top = 4, bottom=12 ( to cut away head shifting noise and black borders )
Filters: Deinterlace = slower, and if needed Denoise = NLMeans, Light
Video: Video Codec = H.264, Quality 20
I tried a bunch of other free and not free converters with terrible results, many just throw away half of the lines or blends them together. I have no experience with Avisynth, so I dont not know if that is better.
Best Regards
J. Nielsen
http://www.datanord.dk/
|
12-03-2016, 09:20 PM
|
|
Premium Member
|
|
Join Date: Aug 2009
Location: N. Carolina and NY, USA
Posts: 3,648
Thanked 1,322 Times in 991 Posts
|
|
Quote:
Originally Posted by jnielsen
I use Handbrake for making deinterlaced MP4 ( from interlaced MPEG2 ) because many computers cannot show MPG or only shows it interlaced.
|
We see this statement often. It's not true. Most PC media players deinterlace by default the same way your external players and TV do (except VLC Player, where you have to set deinterlacing in the options menu). MPEG is one of a handful of universally PV- playable codecs around since the advent of the PC. If your PC can't play MPEG, something's wrong with your PC or your use of it. Of course denterlacing is required for internet posting.
I can't recommend details with Handbrake, which I don't use because of its many limitations, as well as its habit of allowing users to output nonstandard framing formats, among other processing issues and omissions.
Quote:
Originally Posted by jnielsen
I have no experience with Avisynth, so I don't not know if that is better.
|
Much better. Avisynth is an entirely different universe than the software you've mentioned. I should add that NLMeans is one of several versions of several hundred denoisers, repair filters and free external plugins, plus another several hundred more built-in functions, that are available with Avisynth.
|
12-03-2016, 11:07 PM
|
|
Premium Member
|
|
Join Date: Aug 2009
Location: N. Carolina and NY, USA
Posts: 3,648
Thanked 1,322 Times in 991 Posts
|
|
Quote:
Originally Posted by fakir
Okay, so YV12 would be a loss of information? Then we won't need to discuss h264 anymore. (h264 has lossless setting crf=0 which from my understanding should work about like MJPEG, but as it only supports YV12 it's not lossless for me. I wondered if YV12 would only be another way of storing the same information, thus my question.)
|
I'm afraid you have several concepts confused. Capturing PAL to YV12 isn't a matter of loss of data, but a matter of color fidelity and a matter of lossy compression. However, if you would prefer 10-15% percent data loss through compression and 50% less color info to work with in restoration, that's up to you. Lossy is still lossy, even it is high bitrate "near-lossless" material. That might work for ultra-high bitrate massive enlargements used in professional processing software at tens of thousands of dollars USD in cost, but it doesn't fly well for standard definition capture with hobbyist gear. Huffyuv, Lagarith, and UT Video are high-speed codecs designed for real-time lossless capture and playback, which is why they are recommended for capture with a consumer PC.
Quote:
Originally Posted by fakir
I should get a grabber which is capable of capturing 50 frames per seconds to capture every field as half a frame......
So what would be the best way to go on?
|
The most sensible way to capture the tape formats you mention is to copy their frame structure as-is: top field first, interlaced, 25fps. If you want to play with frame structure, do that in post-processing.
Quote:
Originally Posted by fakir
I guess I need to try to set up avisynth through wine and come back with samples, for recommendations for the filter chain?
|
Please post samples as short cuts from original captures, not processed versions. Most lossless samples in these forums are cut with VirtualDub and saved using "direct stream copy" mode to avoid recompression or colorspace conversion. If you don't know how this is done, feel free to ask.
|
All times are GMT -5. The time now is 09:07 PM
|