Are there no good lossless video compression algos?

[CursedLemon]

I was wondering about this, as I've never seemed to encounter them before. Are lossless video compression schemes just not good enough at creating a good compression ratio for video?

[jagabo]

Like with audio, lossless compression algorithms rarely result in much compression. And there never will be. It's just the nature of the beast.

[AlanHK]

If there were, we'd all be using them.

[CursedLemon]

Like with audio, lossless compression algorithms rarely result in much compression. And there never will be. It's just the nature of the beast.

Lossless audio compression in some cases can yield up to a 50% reduction in filesize, as is the case with FLAC. I'm guessing this obviously is not the case with video.

Can anyone explain to me exactly why video cannot be compressed this way? This is an obviously ignorant statement but I just figured that video would be ripe with redundancy.

[poisondeathray]

Lossless audio compression in some cases can yield up to a 50% reduction in filesize, as is the case with FLAC. I'm guessing this obviously is not the case with video.

Can anyone explain to me exactly why video cannot be compressed this way? This is an obviously ignorant statement but I just figured that video would be ripe with redundancy.

Your point of reference is a bit flawed. Your lossess audio example compares uncompressed WAV to lossless compression (FLAC)

The same can be said for video. If you start from uncompressed RGB, you can get 50% reduction in filesize using lossless compression too.

The analogy would be taking a compressed audio file , eg. mp3, and converting to FLAC. So of course the filesize is going to be huge, but much smaller than uncompressed. The same thing happens in video, taking a compressed video file (e.g. h.264, xvid etc...) and converting to lossless compression eg.. lagarith = filesize huge , but much smaller than uncompressed RGB

[jagabo]

Also note that all the high compression lossy codecs use lossless compression (huffman) as the last step in their compression.

[CursedLemon]

Lossless audio compression in some cases can yield up to a 50% reduction in filesize, as is the case with FLAC. I'm guessing this obviously is not the case with video.

Can anyone explain to me exactly why video cannot be compressed this way? This is an obviously ignorant statement but I just figured that video would be ripe with redundancy.

Your point of reference is a bit flawed. Your lossess audio example compares uncompressed WAV to lossless compression (FLAC)

The same can be said for video. If you start from uncompressed RGB, you can get 50% reduction in filesize using lossless compression too.

I'm not sure I understand, that's basically what I was asking (going from uncompressed to lossless). Are there lossless compression schemes for video that have that kind of effectiveness?

[poisondeathray]

Yes, but it also depends on source complexity, much like audio

What is your source format? uncompressed RGB or some other format? and what are your goals (what are you using it for)?

The higher the compression, the less likely you will get real time playback or editing capability in a NLE

There are many different lossless video codecs. You can choose one that offers better compression at the expense of encode or decode speed (tradeoffs)

Lagarith RGB is probably one of the most commonly used because of it's high compression but very poor for decode speed/editing. Huffyuv or huffyuv-mt is less compressed but much faster. UT is even faster, and can offer realtime editing/playback in NLE's on 1080p material with a fast i7

Do a search, as this topic has been addressed before here and on other forums. If you have specific requirements, then ask.

[CursedLemon]

Yes, but it also depends on source complexity, much like audio

What is your source format? uncompressed RGB or some other format? and what are your goals (what are you using it for)?

The higher the compression, the less likely you will get real time playback or editing capability in a NLE

There are many different lossless video codecs. You can choose one that offers better compression at the expense of encode or decode speed (tradeoffs)

Lagarith RGB is probably one of the most commonly used because of it's high compression but very poor for decode speed/editing. Huffyuv or huffyuv-mt is less compressed but much faster. UT is even faster, and can offer realtime editing/playback in NLE's on 1080p material with a fast i7

Do a search, as this topic has been addressed before here and on other forums. If you have specific requirements, then ask.

There's nothing I need to convert, I was just curious. I usually have a lot of questions about video since I find it interesting but I don't know much about it. If I ever were to invest in any video equipment in the future, I suppose it'd be good to know things like this. I guess my hypothetical "purpose" would just be for archiving and storage, in which case I'd take the highest compression ratio possible. Anyway, thanks for answering.

[poisondeathray]

While the conversion from original format to lossless format is "lossless", the conversion back is not.

For archiving/storage - It's usually a better idea to store it in the original format, whatever that is. Chances are it will be smaller in filesize/bitrate (e.g. DVD, blu-ray, camcorder video, virtually anything). The same can be said for lossy audio. e.g. you buy some itunes , it would be better of leaving it as aac or mp3 vs. converting to a lossless format.

It's only the case when your original format is uncompressed, that using a lossless format might might sense (e.g. video capture, CGI renders & animation, etc...)

In video, lossless intermediates usually have a role in intermeditate stages of video editing only. e.g. where you need to import between applications or you have render heavy effects that you want to only render once (vs. at multiple stages)

[CursedLemon]

Well I just found it a little odd, personally. As someone who deals with audio, I pass around uncompressed .wav files all the time. Obviously uncompressed video is going to be a little less practical, but I figured that I might at least HEAR of small project videos being uploaded on the internet fully uncompressed (but in a lossless scheme, of course).

[poisondeathray]

Well uncompressed stereo WAV audio is about 1.536Mbit/s right?

The data rates for 1080p24 uncompressed are like 800-900Mbit/s. Regular hard drives cannot handle it , you need to put them in RAID-0 array. "Passing around" uncompressed HD video for fun is a lot less practical

[CursedLemon]

Well uncompressed stereo WAV audio is about 1.536Mbit/s right?

The data rates for 1080p24 uncompressed are like 800-900Mbit/s. Regular hard drives cannot handle it , you need to put them in RAID-0 array. "Passing around" uncompressed HD video for fun is a lot less practical

Huh, I hadn't even considered that. Makes a little more sense, that.

[jagabo]

For example, In VirtualDub I created a 720x400, 60 fps RGB video from a 1920x1080 60fps source and saved it as uncompressed RGB. The resulting file was 651 MB. Using Lagarith in RGB mode reduced the file to 305 MB. So the compression ratio is similar to the rates you see with losslessly compressed audio.

Note that video generally doesn't have more redundancy than audio when using lossless compression techniques. Even in a still shot with nothing moving and large areas of flat color/shading pixels will be different from their neighbors and from frame to frame because of noise. Lossless compression requires that all the noise be preserved.

[vhelp]

and also..

lossless (in theory) works when the "source" (assuming multiple frames) are in very close relation from one frame to the next, usually in the prev/curr/next (P/C/N) sequence, until a scene-change occurs.

The cleaner the source (based on origin: ie, vhs vs cabtv vs laserdisc vs dvd vs bluray, etc) the better the compression and/or filesize.

what minimizes some of the compression (aka, lossless-ness) is how much PCN differs from each frame, frame to frame. If there are more random pixels (noise or chroma crosstalk) in the images then the compression will not be optimum. The more each frame look a like the better there compression ratio depending on the algo used in the compression in question..there are many different types of compression algorithms.

when people apply a certain type of "filter" chain in their video it is to help to some degree with compressing the images, which also helps in filesize, too..though sometimes at the expense of one aspect of the image/video, aka artifact or other (lossy-ness) error. So some compression may not work in respect to PCN image processing, although I was giving an example of one way of looking at compression from this perspective.

an crude example:

in theory, these numbers would compress more effeciantly
because most numbers are repeated in each of the matrix, temporally.

1 1 1 .. 1 1 1 .. 1 1 1
2 2 1 .. 2 2 1 .. 2 2 2
2 2 2 .. 2 2 1 .. 2 2 1

than say, these, will not because virtually none of the numbers
are repeated in each of the matrix's, temporally or otherwise.

1 1 1 .. 2 1 3 .. 1 4 5
1 5 6 .. 2 3 1 .. 0 1 6
0 6 4 .. 6 1 5 .. 7 6 4

-vhelp 5219

[CursedLemon]

Thanks you guys, very helpful.

[jagabo]

lossless (in theory) works when the "source" (assuming multiple frames) are in very close relation from one frame to the next

Lagarith and HuffYUV are intra-frame only. Most lossless codecs meant for intermediate editing are intra-frame only for speedy access. I believe x264's lossless mode and MSU's lossless codecs are inter-frame.

[El Heggunte]

The VfW version of "Motion PNG" has an inter-frame compression option,
but its author warns that the feature is buggy (and slow as hell ).

[lordsmurf]

While the conversion from original format to lossless format is "lossless", the conversion back is not.

I've been saying this for years, but nobody ever wants to believe me.

[jagabo]

While the conversion from original format to lossless format is "lossless", the conversion back is not.

I've been saying this for years, but nobody ever wants to believe me. :lol:

I didn't understand what he was saying at first. Just to make it clear:

Conversion from a lossy format (like MPEG or DV) to a lossless format is lossless (assuming no colorspace or subsampling conversion). Any program accessing the lossless file will be getting exactly the same data that it would get by accessing the original lossy file directly. But converting that lossless file back to the original lossy format will be lossy. It would be the same, for example, as recompressing an MPEG file to MPEG again.

Converting a lossless format to another lossless format and back is lossless (again, assuming no colorspace or subsampling conversion). You can go back and forth a hundred times and the final data will be exactly the same as the original.

[lordsmurf]

That's not entirely true either. The various "lossless" don't have the same colorspace conversion.

There are definite losses between something like HuffYUV and uncompressed YUY2, but I really have to search in high action scenes for it. It is near-invisible.

A lot of people get confused over what lossless means, why it exists, etc. You know. I know. Poison knows.

[jagabo]

The various "lossless" don't have the same colorspace conversion.

That's why I said "assuming no colorspace or subsampling conversions".

There are definite losses between something like HuffYUV and uncompressed YUY2

When done correctly there are no differences.

but I really have to search in high action scenes for it.

HuffYUV is an intra-frame only codec. It makes no difference if there is motion between frames or not.

Poison knows.

I'll let him speak for himself.