SPIHT wavelet codec and another...

[vsv]

Where's possible to download this codec?
What do you think about these new wavelet "visually lossless" 10 bit 4:2:2 yuv codecs as Cineform ProspectHD 10 bit, Avid DNxHD and SPIHT ?
These codecs good solution for new generation tapeless HDcam.
Ikegami have one: http://www.ikegami.com/Avid_NLTek.html
Anybody have information about implementation SPIHT in FPGA design?

[Zetas]

Cineform & Avid DNxHD
http://www.abcdefg.com.ua/

EDIT by Koepi: No warez! Those codecs aren't free!

[video]

avid's DNxHD is almost free (http://www.avid.com/DNxHD/codec/eula.asp) ;)

[vsv]

Originally posted by Zetas
Cineform & Avid DNxHD
http://www.abcdefg.com.ua/

EDIT by Koepi: No warez! Those codecs aren't free!

Alf, what you can say about SPIHT? CFHD no doubt very good!

[Sharktooth]

w00t? 116mb for a codec?!?! Ppl at AVID is crazy...

[communist]

Its the source code :rolleyes: ;)

[Leak]

Originally posted by communist
Its the source code :rolleyes: ;)

w00t? 116mb for a codec's source code?!?! Ppl at AVID is crazy...

Think about it - even the linux kernel sources are less than a third of that in size, and that's a whole OS kernel with drivers for all kinds of stuff thrown in already...

np: Markus Guentner - Hooked On Monkey Phonics (ft. Mario Thümmler) (Audio Island)

[communist]

Oh well there goes my lame attempt at explaining the size.. well the reason for the big size after all seems to be:
from the README.TXT
The source code also contains a set of test frames that can be used to verify the proper functioning of the codec. The test frames are stored in the TestFrames directory. There are a set of reference, uncompressed 1920x1080 frames in the UncompressedFrame1920_1080 sub-directory and a set of reference, uncompressed 1280x720 frames in the UncompressedFrames1280_720 sub-directory. These sub-directories also hold .TIFF files for reference.
Happy downloading ;)

[akupenguin]

2D SPIHT is a very good still image compressor (both for lossy and lossless). 3D SPIHT sucks for video coding, because it lacks motion compensation. (All the tests that say otherwise just happen ;) to use very low motion videos as test subjects.) There are ways to add motion compensation to a 3D wavelet codec, and I'm sure those would make SPIHT competitive, but AFAIK they haven't been combined yet.

[Leak]

Originally posted by akupenguin
2D SPIHT is a very good still image compressor (both for lossy and lossless). 3D SPIHT sucks for video coding, because it lacks motion compensation.

Isn't the lack motion compensation rather negligible for a lossless codec that's meant for editing, though? After all, what you want from it is being able to jump to every frame, and fast at that...

No objections about motion compensation being a must for a format meant for end user playback, of course...

np: Rhythm & Sound with Rod Of Iron - Lightning Storm (See Mi Yah)

[akupenguin]

3D SPIHT is temporally predicted. Not exactly with I and P frames like you're used to, but you still can't just seek to any frame. It's also not fast to decode.
2D lossless SPIHT as an intra-only editing format could be interesting, though.

[vsv]

akupenguin, thank you for information about SPIHT.
I've did quick test "CineformHD vs Avid DNxHD" by method
described on http://codecs.onerivermedia.com/ and can say:
CFHD not freeware, but is my choice - just amazing codec!
You can download trials from www.cineform.com
ConnectHD for Vegas and AspectHD for Premiere.
Avid DNxHD codec ask from Zetas.

[vsv]

Mystery Science Codec 3000, a fully working wavelet-based video/image codec with some interesting features... (http://www.moosehill.se/index.php?page=msc3k/msc3k.php)

MSC3K is developed by Jonas Nilsson

http://www.moosehill.se/msc3k/gandalf.png

Features of MSC3K
Good compression performance at low bitrates!
Still image compression.
Video compression (P-frames only) with motion estimation.
Fully scalable quality at bit-level for still images!
Scalable quality for video, perfect for streaming!
Really fast encoding.

[Teegedeck]

Well, if that MSC3K ain't vapor I'd say, get the man and put him into the SNOW development department! :)

[iwod]

Originally posted by Teegedeck
Well, if that MSC3K ain't vapor I'd say, get the man and put him into the SNOW development department! :)

Agree

[vsv]

Originally posted by Teegedeck
Well, if that MSC3K ain't vapor I'd say, get the man and put him into the SNOW development department! :)

Just few words from Jonas Nilsson :
"Unfortunately I'm not currently developing the codec anymore so this page is actually a cry for help! I lack the inspiration and the determination to complete my work on MSC3K, but maybe if YOU would drop me a mail and say what you think of the results here I would find some inspiration =)
What do you think? Please let me know!
MSC3K is developed by Jonas Nilsson just to make the world a better place for all of us. Cheer me up and make me finish this work! Drop me a mail! " http://www.moosehill.se/index.php?page=msc3k/msc3k.php

[video_magic]

I visited his page just now. The work looks great and I emailed him to let him know this.

[Kefir]

Hello

I got an email from one of you who had seen my page. It took me 5 days before i could post a reply though. If you are interested in what i coded on (MSC3k) i'll happily provide you with some anwsers.

I know most of the theories and maths behind the wavelet transform and the arithmetic compression schemes so if someone wants me on some other project i might find some time for that.

/cheers
Jonas Nilsson

[ac-chan123]

Is there any way to get the Codec? I will test it with some VQEG Sequenzes. Videocodec can't be compared by picture, they must be compared by sequenzes.

Some questions i have allready:
-is it a Keyframe only Codec like MJPEG
-wich restriction are for the resolution(multiple of some think)

[Kefir]

The codec isnt completley finished. I stopped coding on it 2 years ago. It has I and motionestimated P frames only and should be a multiple of 16pixels preferably 32 pixels.

/Jonas

[ac-chan123]

Is there any reason to don't public the code?

[Ark]

Kefir, thanks for finding some time to reply to our questions, it's always good to have direct contact with developers :).

As Teegedeck pointed out, you'll surely be a valid contributor for the SNOW project, which is slowly proceeding (afaik) lately.

[Kefir]

Apart from being far from finished with alot of things that I should have fixed a long time ago, no , I guess not.

/Jonas

[PatchWorKs]

...so let "us" help you ! ;)

[video_magic]

Hello, and I'm glad you visited :)

Here is a link to the person that writes the famous BT WDM opensource drivers for the TV capture cards. I have just noticed this from 2003 which says he has written a wavelet codec
http://sourceforge.net/forum/forum.php?forum_id=262721

"Is there any interest on an OpenSource realtime codec based on the lifting Daubechies9/7 Wavelet decomposition (floating point)? - I have written such codec, it runs in realtime at 720x576..."