Capturing Laserdisc

[gshelley61]

Laserdiscs are uncompressed digital video. That is why they are so big. They also are in 4:2:2 colorspace, which would benefit from MJPEG compression, not DV (or MPEG which is worse than DV). I miss MJPEG. If you truly want to archive your Laserdiscs, then capture in MJPEG. Later you can screw it all up by compressing it to MPEG2 for DVD viewing.

Actually, laserdiscs are not digital video. They are analog and therefore don't have a sampling colorspace. This is why worm noise and other laserdisc artifacts are visible when playing them back. The physical medium has an effect on the quality, unlike digital where the medium is irrelevant. Digital PCM audio tracks were eventually added to laserdiscs, but the video itself is plain old composite analog.

More detailed info on the LD format:

http://en.wikipedia.org/wiki/Laserdisc

[tedkunich]

@tedkunich, two things, one: http://www.pegasusimaging.com/picvideolossless.htm
That's from the JPG webpage you quoted me, using the links from the pulldown menu, so I really don't see how you didn't see anywhere that says Lossless

My bad... did not see that one... guess I'm blind as well as deaf!

Secondly, allow me to quote you now ok ? "If you compare every bit of the HuffyuvBMP file with the original, they will be identical" - If I use Huffyuv on an Uncompressed AVI (no compression of any kind) you still claim it will be identical ? cause if not, then it's also a compression of different kind. and as long as it has compression in it, you loss data.

What I was trying to say:

Raw AVI -> Vdub -> save frame x as BMP1
Raw AVI -> Huffyuv encode -> file1 ..... file1 -> Huffyuv decode ->Vdub -> same frame x as BMP2
Raw AVI ->MJPEG encode -> file2 ..... file2 -> MJPEG decode ->Vdub -> same frame x as BMP3

Performing a bit compare on BMP1 and BMP2 should be an exact match. Comparing BMP1 and BMP3 would not be a exact match, assuming that the codec was indeed lossy. As I put my foot in my mouth too soon, this whole excersise becomes moot.

It's not a "pissing" contest, it's just to prove that the debate of Lossless MJPEG compared to Huffyuv is not relevent. Each compresses differently, you WILL loss data compared to Uncompressed AVI.

I think one thing that is lost in the discussion of lossy vs lossless is the matter of colorspace. Converting beween the various color spaces, RGB, YUV, 4:2:1, 4:2:2, etc, is not without error and one can expect different results with different codecs depending on source and destination colorspaces and tools used.


T

[lordsmurf]

Actually, MPEG2 is better than DV (NTSC).
4:2:0 is better than 4:1:1

DV is okay as a shooting format, but the "moaning" from non-consumers is still quite loud.

[junkmalle]

HuffYUV is a completely lossless codec. What this means is that the YUY2 data that comes out after decompression is exactly the same as the YUY2 data that went in before compression.

You can argue that the colorspace conversion that takes place in converting RGB to YUY2 is lossy. There are two conversions involved. In the first step RGB is converted to YUV (essentially one byte of intensity information and two bytes of color information) like this RGB RGB RGB RGB --> YUV YUV YUV YUV. There is a small loss of accuracy in this step because there isn't a 1 to 1 correspondance between the RGB and YUV colorspaces. In the second step half the color information is thrown out. Instead of having YUV tripples you are left with pairs with Y's alternating U and V values: YU YV YU YV. So the colorspace conversion loses a little accuracy in the RGB to YUV conversion and then each pixel is reduced from 3 bytes to 2 in an obviously lossy process.

But before you go getting all excited about how MJPEG is so superior -- the first step of JPEG compression is to convert from the RGB colorspace to YUV colorspace and to reduce the number of color constituents -- ie, it's more or less the same as YUY2! And it's basically the same for MPEG, Divx, and all the high compression codecs. NTSC video is very close to YUY2 in that it carries less color information than it does intensity information. So all capture cards capture that signal in something like YUY2 and then convert it to RGB to be saved to disk.

HuffYUV does have a slightly lossy YUV option but it's not the default. It can also work losslessly in an RGBA colorspace. Look at the options in the codec's configuration dialog.

PicVideo has a lossless MJPEG codec but that is not the codec most people refer to when they are compressing video. Most people use the lossy version of the codec because it gives much higher compression rates. The lossless version gives compression ratios very similar to HuffYUV.

The problem with all the lossless codecs is that you can't compress real world video with them very much. You're lucky to get a 2:1 compression on most video.

[DTSL06]

Perhaps I'd be better off capping to my DV-m10s and then ripping the disc to the PC for authoring. I just hate to bypass all this expensive PC equip for a $350 solution I bought on Amazon.

Thats how I do my LD transfers, also my VHS transfers too, using my JVC DR-M10.

[Sefy]

@junkmalle, now hold everything! lets go back to that little thing you said that will make this whole "debate" a lot of air over nothing, quoting you here: "PicVideo has a lossless MJPEG codec"

IS IT or is it NOT a Lossless MJPEG ? cause you'd be very surprised, but that's the codec i used myself, if you or others don't, does not mean anything, the fact it DOES exist is the whole point of this debate.

[SatStorm]

I was lucky enough to download PicVideo before turned commercial. It was freeware once and I have it from back then. I never upgrade to newier versions.

I too prefer it over Huffyuv. In practice, I saw better results with picvideo than Huffyuv.
Maybe, this happens because I don't use much CCE. I mostly use TMPGenc all those years. And as we know, TMPGenc convert YUV2 to RGB and back to YUV2 to encode it to mpeg 2. CCE handles direct YUV2 and maybe that is what makes the difference.
After years of testing, it seems to me that TMPGenc handle better PicVideo than Huffyuv.

Just for the info, there is also a freeware variaton of picVideo: It is the Asus video codec ver 2. I don't know if it is possible to use it on winxp or even win2k (I believe it is only vfm), but the results when I use to use it, was almost like picVideo and still far better Huffyuv (using TMPGenc to encode)

Also, I believe all this has to do and with the resizing:
IMO, you can resize the mjpeg based codecs and have better results than Huffyuv codec. Why resizing is important, you might ask:
Well, as you all probably know, the capture cards capture at a specific framesize and then they resize to what we ask them the first place.
For example, the ATI cards captures at 704 x 576/480 framesize. When we set to capture at 720 x 576, the ATI card gonna capture at 704 x 576 and then, using internal rutines, gonna resize to 720 x 576 on the fly. (BTW this was the reason that earlier ATI cards sucked in quality for us, the PAL users. All the PAL captures, use to be a product of a 688 x 480 framesize if I remember correct because of the stupid ATI drivers. New models don't do this anymore)

Anyway, why I mention all this:
Because when you know what is the native framesize of your capture card, you can use that specific framesize to capture with the Huffyuv codec, that way you can have top results.
But if you don't know the native framesize of your card, then the internal resizing which your card must do to reach what you set the first place, can destroy the benefits Huffyuv gave you.
When you use picvideo, you have a better final product regarding resizing. It keeps better sharpness, has better colours and less "fuss" huffyuv.

This is my practical observation all over the years.

[junkmalle]

The lossless codec from Pegasus Imaging is called "PICVideo Lossless JPEG codec". The lossy version is called "PICVideo MJPEG codec".

[Sefy]

@junkmalle, but I thought there is no such thing as a Lossless Compression if it has any PEG technology in it

[junkmalle]

@junkmalle, but I thought there is no such thing as a Lossless Compression if it has any PEG technology in it

The Joint Photographic Experts Group also defined a lossless and near-lossless compression scheme. It was based on HP's LOCO codec. I assume Pagasus (PICVideo) uses this algorithm because HP offers it royalty free.

[Sefy]

@junkmalle, but people here insisted there is no Lossless anything when it comes to PEG type of compression, so someone is gonna have to make a stand. Is there or is there NOT a lossless PEG ?

Cause as far as I can see by logic, if you have Lossless JPEG, there is no reason why there won't be a Lossless Motion JPEG which would be the same only for moving images

[junkmalle]

The PICVideo Lossless JPEG Codec is a lossless JPEG codec for video -- ie, it is a lossless MJPEG codec. I suspect very few people use this codec because, like all lossless codecs, it doesn't compress real world video very well. Most of them use the lossy PICVideo MJPEG codec.

[junkmalle]

Here are some numbers for anyone who's interested. I took a 720x480, 30 fps, progressive, uncompressed (originally DV), ~30 second video and compressed it with HuffYUV, PICVideo MJPEG (4:2:2 sampling), and PICVideo Lossless JPEG codecs. The files sizes:

Uncompressed RGB: 908 MB
Lossless JPEG: 639 MB
HuffYUV: 227 MB
MJPEG (20): 237 MB
MJPEG (19): 94 MB
MJPEG (18): 62 MB

I also verified that the PICVideo Lossless JPEG compression is completely lossless (at the lossless setting, of course). The output was byte-for-byte identical to the input.

[tedkunich]

Here are some numbers for anyone who's interested. I took a 720x480, 30 fps, progressive, uncompressed (originally DV), ~30 second video and compressed it with HuffYUV, PICVideo MJPEG (4:2:2 sampling), and PICVideo Lossless JPEG codecs. The files sizes:

Uncompressed RGB: 908 MB
Lossless JPEG: 639 MB
HuffYUV: 227 MB
MJPEG (20): 237 MB
MJPEG (19): 94 MB
MJPEG (18): 62 MB

I also verified that the PICVideo Lossless JPEG compression is completely lossless (at the lossless setting, of course). The output was byte-for-byte identical to the input.

Are you sure that you had Huffyuv set correctly? I only see about a 1.8:1 compression at best with Huffyuv, never ANYTHING close to the 3.8 you have. What tool was used to test the codecs, VDub?


T

[junkmalle]

That particular video does seem to be more compressible than usual with HuffYUV. Actually it's even worse than you calculated 908/227 = 4.0! But are you comparing to uncompressed YUY2 or uncompressed RGB? YUY2 will be 2/3 the size of RGB to start with. If compared to uncompressed YUY2 the compression ratio is about 2.67 to 1.

I checked all my numbers (I haven't deleted the files yet) and I didn't make any typos. I also looked closely at the HuffYUV compressed file and it looks fine. All of the files have the same frame rate, frame size, no audio.

[tedkunich]

But are you comparing to uncompressed YUY2 or uncompressed RGB? YUY2 will be 2/3 the size of RGB to start with. If compared to uncompressed YUY2 the compression ratio is about 2.67 to 1.

Doh! I'm so used to dealing with YUV that I forgot about that....


T

[Sefy]

So to make a long story short, you just proved that there is such a thing as a Lossless JPEG codec, so this entire debate is basicly now over