Announcement: New open source DV codec (VfW) [Archive] - Page 3

IanB

21st November 2006, 02:07

It is not the codec that chooses the output format. It is the application you are using.The rule YV12 -> YUY2 -> RGB32 -> RGB24 applies only to AviSynth's AviSource()! Other applications have other rules that suit their needs.

AviSynth's DirectShowSource() accepts the very first format of {RGB24, RGB32, ARGB, YUY2 and YV12} offered by DirectShow. It does not try before it buys ;)

When using AviSynth's AviSource() or DirectShowSource(), the easiest way to get a codec to decode into the format of your choice is to set Pixel_Type="..." appropriatly. Doing this will cause ...Source() to negotiate for the chosen format only and throw an error if the codec cannot provide this format. i.e. 1. Asking the Panasonic codec for anything other than RGB24 will always fail. 2. Asking Cedocida when it is in forced output mode for any other format will likewise fail.

For other applications that do not have a "Pixel_Type" functionality then you must use the option in Cedocida to only connect with the forced format. Of course the application has to be willing to work with the forced format. i.e. Old TMPGEnc versions will only work with RGB24.if i select force output and chose yuy2 or if i chose yv12. In both case i got yuy2.For NTSC there is no YV12 output, so YUY2 appears to be subtituted, this is probably a bug.

Serbianboss

21st November 2006, 10:07

@Fizick

Any other server for download, this is so slow,so i cant download file.

thanks

Fizick

21st November 2006, 18:43

Unfortunately all drives of RapidShare.de are full right now.

Wilbert

21st November 2006, 19:23

Serbianboss

21st November 2006, 21:47

This server is lot better.Thanks

@Fizick

Did you change the code of cedocida? I see that you make boxes instead radio buttons.

I found interesting things.

1. When in decoder option i select RGB,YUY2 and YV12(all option is selected when press default button) my dv avi files are in YV12 colorspace. And dv avi can load in virtual dub

2. But, when in decoder option i just select YUY2, my dv avi files are in YUY2 color space. But problem stay the same. Virtual dub cannot load dv avi files, but it can load if we make simple avisynth script.

Fizick

21st November 2006, 22:03

Serbianboss,
select yuy2 and rgb :)

Serbianboss

22nd November 2006, 01:57

That's work

Just to ask. When is selected RGB and YUY2 now i can load Dv avi in virtual dub.
I am now interesting how in virtual dub sais that my source is yuy2 when i checked yuy2 and rgb colorspace.

Here is picture:
http://img90.imageshack.us/img90/8682/untitledab1.jpg (http://imageshack.us)

I mean everything for now working ok, but i am interesting about this.

2. One more question: When load simple avisynth script

AVISource("C:\\Desktop\dv avi movie.avi")
info()

in virtual dub said that my Decompressor is YUV 4:2:2 and my FourCC is also YUV 4:2:2. Is that Ok or no? I am in pal country shuldnt be 4:2:0 ?

thanks

Serbianboss

22nd November 2006, 22:16

any comment about this

Wilbert

22nd November 2006, 23:30

If you don't select YV12 in 'Enabled output formats' it won't output its native color format YV12 (for PAL).

AVISource("C:\\Desktop\dv avi movie.avi")
info()

in virtual dub said that my Decompressor is YUV 4:2:2 and my FourCC is also YUV 4:2:2. Is that Ok or no? I am in pal country shuldnt be 4:2:0 ?
Yes, that's ok. As you can see in IanB's post, from the enabled formats, AviSynth tries to get YUY2 first from the decoder (and if that's not possible it receives RGB).

Just to ask. When is selected RGB and YUY2 now i can load Dv avi in virtual dub.
I am now interesting how in virtual dub sais that my source is yuy2 when i checked yuy2 and rgb colorspace.
Because (1) YV12 is not enabled in the decoder and (2) you selected 'fast recompress' (that's a hunch, so correct me if i'm wrong) in VDub. If you select 'full processing mode' in VDub, the DV decoder should give you RGB stuff.

Serbianboss

23rd November 2006, 00:06

Thanks for answer.

You are right. When in virtual dub is selected full compression it output RGB,when is fast recompress it output YUY2.

This is for virtual dub, but what you think for CCE. When in cedocida decoder is selected YUY2 and RGB what colorspace then output CCE ?

cedocida

27th December 2006, 15:52

Hi Andreas,
There is a problem in the normal quality encoder too. I noticed it when doing a 90-minute PAL encode. The first hour encoded great, and then suddenly the strange violet-gray blocks started popping up randomly. It's probably the same bug as in the High Quality mode, but just less obvious.

Currently I am working on bug-fixing and concerning the issue with the corrupt blocks using high quality mode, I can reproduce it and I do understand the reason for it.
The problem with normal quality mode, reported above, needs further investigation. For that, @zambelli:
can you upload the frames where the encoder produces this strange blocks?

@Fizick - your suggestion about the GUI:
Thank you for your suggestions. At the moment I want to concentrate only on bug fixing and most probable I will not touch the GUI in the future. Anyhow your modifications are welcome - the project therefore is open source.

Regards,
Andreas

stegre

2nd January 2007, 07:33

A. Dittrich -

If I'm not mistaken, I think there is a significant bug in the DV file structure produced by cedocida. I was working on bugs in GSpot's handling of DV, some of which led me with questions which remain frustratingly unresolved (I may post a separate thread to see if I can get some help here on those issues). Meanwhile, in trying to resolve those, I compiled your encoder figuring it would be useful to be able to create various test files, especially by having the encoder source to modify for experiments.

In so doing, however, I was surprised to see that GSpot was rejecting the stock cedocida files altogether (though they do seem to play fine, somewhat surprisingly to me).

Anyway, to get to the point, see screenshot below, with table from SMPTE-314M, which I'm sure is the same as its counterpart in IEC 61834. The VS and VSC pack data is supposed to go in frames 0 and 1 on odd DIF sequences, and 39 and 40 on even ones. I believe your code is putting them in completely the wrong place on the odd sequences (like sequences 29 and 30, or 30 and 31, or something), probably overwriting something that should be there - and leaving it missing from where it should be. I tracked it down to the line below in your source (encode.cpp, line 335 in the 0.1.6 source I have). I think a correct fix is just to change the single multiplier as shown below. Please let me know if you concur, or perhaps I'm off-track somehow.

http://ftyps.com/unrelated/cedocida.png

cedocida

3rd January 2007, 00:01

@stegre:
yes, you are right in both: finding and solving the bug.
It is now fixed in the

*** released new version 0.1.7 ***

available here (source code only):
www-user.rhrk.uni-kl.de/~dittrich/cedocida (http://www-user.rhrk.uni-kl.de/~dittrich/cedocida/index.html)

changes:
- Fixed: bug in encoding/decoding right border of interlaced ntsc video
- Fixed: bug in encoding if frame activity is too high
- Fixed: bug in DIF-Header, wrong position of video-packets
- Fixed: minor bugs in GUI

Guest

3rd January 2007, 00:06

I will update my hosting with the new binary tonight at neuron2.net.

Guest

3rd January 2007, 02:42

Version 0.1.7 binary now available here:

http://neuron2.net/misc/cedocida017.zip

DSP8000

3rd January 2007, 04:51

Tnx. for the new binary neuron2 ;) .

The read me file included it is a bit confusing for me :o ,
For multiple edit of PAL DV video which options in in/output should we choose?

Example:
1.Transfer DV from cam,
2.Edit in any NLE,
3.Export the timeline to Cedocida codec.

Now, what options should we choose in the enc/decoder in order to avoid colorspace conversion & get proper sampling of the video?

Most of the NLE apps do internal upconversion to 4:4:4 when we apply transitions, color corrections, video filters, etc...
Most of them require RGB input.

My questions are these:
1.For encoding to MPEG2 interlaced what do we choose in the codec options enc/dec?
2.For encoding to Cedocida codec for multiple edit or transfer what do we choose in the codec options enc/dec?
3.For proper decoding of PAL DV video what do we choose in the codec options enc/dec?

DSP8000

Guest

3rd January 2007, 05:05

Give all credit to Andreas! It's not hard for me to make a binary. :)

I'll defer to Andreas for the technical questions on his work.

stegre

3rd January 2007, 16:28

Well, thx to all. File structure looks perfect now.

Chainmax

3rd January 2007, 16:56

I've been doing some DV-to-DVD conversions lately. Most of the source footage I get is deinterlaced PAL, but I'd like to know what are the Cedocida recommended settings use on all cases as I've seen some stairstepping on reds and I might have to deal with interlaced/progressive PAL/NTSC footage soon anyway. The reason I ask is because sifting through this thread hasn't revealed e settings a majority agrees upon.

Fizick

3rd January 2007, 18:53

now we have two v0.1.7 versions. Sorry

Chainmax

4th January 2007, 03:13

I have the one neuron2 released.

Guest

4th January 2007, 04:46

now we have two v0.1.7 versions. Sorry My policy here is to use Andreas' version numbering because I am just building his source (except for correcting one misspelling on the GUI, which I'm sure he won't object to). If you make a derived version with new functionality, you should rename it, e.g., like TDecimate(). Maybe yours could be Fcedocida, or something like that.

Fizick

4th January 2007, 14:24

I hoped that my suggestions be accepted.
I will built some modded version for private using.

Wilbert,
please remove link to my v0.1.7 in your post above to prevent mess.

Serbianboss

4th January 2007, 14:43

I think that is better way that we could in decoder option to check more colorspace option(for example yuy2 and rgb) instead just one

Serbianboss

4th January 2007, 23:01

I found some bug.

If we set SetMTmode(2) at begining of script and using cedocida dv codec and if we want to preview in virtual dub or in avsp i had this:

http://img168.imageshack.us/img168/5307/untitledrr9.jpg (http://imageshack.us)

But if using other dv codec(panasonic) everything is ok.

kxproject

7th January 2007, 21:30

I really like the Celocida DV codec but I have one program that I'm having a problem using it with. In VirtualDub (I’ve tried various versions) if I save a file using Full processing mode, and the Celocida DV codec as the compression type, when I play back the saved DV file in a media player the resulting video is corrupt with large color blocks. This problem does not occur using other DV codecs as the compression type. I’m just wondering is anyone having this problem in VirtualDub?

i have the same problem, i have a green bar at the bottom

cedocida

8th January 2007, 22:50

@kxproject
Does this happen with the current version 0.1.7?

kxproject

9th January 2007, 10:46

@kxproject
Does this happen with the current version 0.1.7?

Yes, both with 0.1.7 and 0.1.6. I had to convert the colorspace to yv12 in avithynth in order to get rid of that green line at the bottom, so it was not the problem specific to cedocida. The same line appeared also with main concept dv codec until i made this conversion. So i really don't know what the cause of the problem. The video file was the usual dv file created by panasonic gs-400 camera. It was rendered as uncompressed in sony vegas after some changing of brightness and contrast and the uncompressed file was opened in Vdubmod via avisynth, converted to yuy2 for applying greedyhma which requires yuy2 and then saved from vdmod in the direct stream copy mode. The options in the cedocida were input = yuy2, output = yuy2.
By the way has anyone compared cedocida with matrox soft dv codec and canopus soft dv codecs? I don't have them but some people swear they are the best.

Fizick

22nd January 2007, 23:29

It seems, Cedocida encoder not properly processes thin color horizontal lines in PAL mode (NTSC normal).
http://forum.ixbt.com/topic.cgi?id=29:22850-6#169
May be wrong YUY2 to yuv420 internal conversion ?

cedocida

23rd January 2007, 20:53

@Fizick:
Can you give a short summary of this issue in english (or german), so if this is really a bug, we can fix it.
Thx in advance!

Fizick

24th January 2007, 06:58

KruFFT is still waiting 5 days to post to doom9 forum.
Here is my analisys:
Cedocida does not filter PAL chroma YUY2 source vertically before conversion to YUV420.
So, colors of some thin lines are simple discarded, see pictures at russian forum.

It may be fixed with avisynth pre-filter script:

function PreFilter420(clip clipYUY2) {
# YUY2 chroma preliminary lowpass filter for PAL Cedocida DV encoder (v0.16, 0.17)
# Fizick, 2007
SeparateFields(clipYUY2)
u=UtoY().Blur(0,1)
v=VtoY().Blur(0,1)
MergeChroma(last,YtoUV(u,v))
weave
}

ImageSource ("post0000.png",fps=25)
addborders(0,0,440,396)
converttoyuy2()
prefilter420()

Blur(0,1) is vertical filter 1/4, 1/2, 1/4 for same fields chroma lines.

krieger2005

24th January 2007, 15:31

A short extension:
1. The first picture is the original source
2. The second picture demonstrate Cedocida
3. The third one MainConcept DV

A next test: http://forum.ixbt.com/topic.cgi?id=29:22850-6#175
left Picture: original
center: MainConcept
right Cedocida
Picture is zoomed 4-times

morsa

4th February 2007, 00:32

Wouldn't sound like a next logical step to add DV50 and DV100 to Cedocida?

I read SMPTE 370M, and it doesn't seem a lot different.

http://www.smpte.org/smpte_store/standards/pdf/s370m.pdf

piscator

24th May 2007, 09:31

I recently upgraded to Windows x64 and the installation of the codec does not work.

Steps:
1) I used the 017 version from neuron2's site.
2) Unpacked and clicked the inf.
3) Got the usual message that it was not a verified driver or some such thing. Pressed continue to proceed.
4) Looked with VCSwap and the codec does not show up (and obviously no dvds content can be played back).

Remarks:
* Got no error messages whatssoever.
* the dll and the inf file are installed in C:\Windows\System32 and C:\Windows\Inf
* I'm not sure but I think i used to have 2 dialogs during driver installs. Now, it just was the message with verified/signed drivers.
* I had it working before on a 32-bit Windows.

So, any ideas how I might get it to work?

thx,
Piscator

cedocida

24th May 2007, 18:39

Most probable this is because of the inf file which is not ready for x64 systems.
As long as no update for the inf file is available, you can try this workarround (see http://www.microsoft.com/whdc/driver/install/64INF_reqs.mspx):

Turn off the undecorated models check for AMD64. To turn off the undecorated models check for AMD64-based systems only, create the following non-zero numeric registry value:

HKLM\Software\Microsoft\Windows\CurrentVersion\Setup:
REG_DWORD: DisableDecoratedModelsRequirement

Set the key to a non-zero value (such as 1) and restart the system, then install the driver.

piscator

24th May 2007, 21:08

Always charming those hidden registry options. It would have even be more charming if it would have worked...

Anyway, I dug a bit deeper.
When I go to Control Panel/Sounds and Audio Devices/Hardware/Video Codecs and properties. It shows the cedocida codec. However, status = "Driver is enabled but not functioning properly". It also shows two radio buttons but with no text.

Against knowing better, I tried to change the INF file and decorated it (as in the article). But I get the feeling that the following part of the article cannot be ignored...
Ensure that drivers and installation routines developed for x64- and Itanium-based systems include 64-bit kernel-mode drivers and 64-bit class installers and coinstallers, as described in this article and in the Windows DDK.

So I guess I'm at a loss for the moment...

greetz,
Piscator

piscator

24th May 2007, 21:19

Found a solution! :-) The following inf file I got from http://forum.videohelp.com/topic316263.html#1622413 did the trick.

; Cedocida DV Codec install file - 15dec06
; Usage: right-click on this file and choose "Install"
; For 32-bit codec on WindowsXP 64-bit Edition

[version]
signature="$Windows NT$"
Class = MEDIA

[DefaultInstall]
CopyFiles=DVR.Files.Inf,DVR.Files.Dll
AddReg=DVR.Add.Reg

[DefaultUnInstall]
DelFiles=DVR.Files.Dll,DVR.Files.Inf,DVR.Files.Ini
DelReg=DVR.Del.Reg

[SourceDisksNames]
1="Cedocida DV Codec","",1

[SourceDisksFiles]
cedocida_x6432.inf=1
cedocida.dll=1

[DestinationDirs]
DVR.Files.Inf=17
DVR.Files.Dll=16425
DVR.Files.Ini=25

[DVR.Files.Inf]
cedocida_x6432.inf

[DVR.Files.Dll]
cedocida.dll

[DVR.Files.Ini]
cedocida.ini

[DVR.Add.Reg]
HKLM,SYSTEM\CurrentControlSet\Control\MediaResources\icm\vidc.dvsd,Description,,"Cedocida DV Codec"
HKLM,SYSTEM\CurrentControlSet\Control\MediaResources\icm\vidc.dvsd,Driver,,"cedocida.dll"
HKLM,SYSTEM\CurrentControlSet\Control\MediaResources\icm\vidc.dvsd,FriendlyName,,"Cedocida DV Codec"
HKLM,"Software\Wow6432Node\Microsoft\Windows NT\CurrentVersion\drivers.desc",cedocida.dll,,"Cedocida DV Codec"
HKLM,"Software\Wow6432Node\Microsoft\Windows NT\CurrentVersion\Drivers32",vidc.dvsd,,"cedocida.dll"

HKLM,Software\Wow6432Node\Microsoft\Windows\CurrentVersion\Uninstall\Cedocida,DisplayName,,"Cedocida DV Codec 32-bit (Remove Only)"
HKLM,Software\Wow6432Node\Microsoft\Windows\CurrentVersion\Uninstall\Cedocida,UninstallString,,"%16425%\rundll32.exe setupapi,InstallHinfSection DefaultUninstall 132 %17%\cedocida_x6432.inf"

[DVR.Del.Reg]
HKLM,SYSTEM\CurrentControlSet\Control\MediaResources\icm\vidc.dvsd
HKLM,"Software\Wow6432Node\Microsoft\Windows NT\CurrentVersion\drivers.desc","cedocida.dll",,""
HKLM,"Software\Wow6432Node\Microsoft\Windows NT\CurrentVersion\drivers32","vidc.dvsd",,""

HKLM,Software\Wow6432Node\Microsoft\Windows\CurrentVersion\Uninstall\Cedocida

zambelli

7th August 2007, 06:26

cedocida

14th August 2007, 22:13

I've run into another encoding bug, this time in "Normal quality" NTSC mode. Build v0.1.7 from neuron2.

Is anybody else seeing issues in the encoder? Or is nobody using Cedocida for encoding?

I am interested in fixing this bug. As there is now a new version out (see update of the first post here) with some bigger changes can you check again and/or post a scenario for reproducing this issue?

Regards, Andreas

cedocida

14th August 2007, 22:31

Other suggestion: may be it will be beter to use checkboxes for enabled video formats instead of radiobuttons with single out format.
(really, I created this code modfication for myself and can provide a sources :) )
I prefer to use YUY2 and RGB switched ON.
Fizick, thanks again for your suggestion (from last year). The latest version 0.2.0 now uses checkboxes for enabling the color formats.

Guest

14th August 2007, 22:56

The build is here (temporarily):

http://neuron2.net/guest/cedocida020.zip

I'll update my web site link tonight.

zambelli

15th August 2007, 09:44

I am interested in fixing this bug. As there is now a new version out (see update of the first post here) with some bigger changes can you check again and/or post a scenario for reproducing this issue?

Thanks, Andreas. I'll try this out sometime this week. If I find encoding errors again, I'll upload some samples.

plane

18th August 2007, 07:14

I want to know why the older build of 0.17 got a Encoder High quality option while the recent 0.17 and 2.0 the highest option available is just Normal quality.

cedocida

18th August 2007, 10:37

I want to know why the older build of 0.17 got a Encoder High quality option while the recent 0.17 and 2.0 the highest option available is just Normal quality.
The old options were mainly used (by me) to test different strategies and comparing encoding speed vs. quality. As it turned out that the old option for the best quality was faster than the old normal option, I decided to rename the old best quality option to "normal" and drop the other options, but not the fast option which is faster but worse in quality.
So, with the "just Normal quality" in 0.1.7/0.2.0 you get the same high quality as in build 0.1.6.

Fizick

18th August 2007, 17:24

cedocida (you change a nick?),

What about my other suggestion: internal low pass prefilter (as an option) ?
http://forum.doom9.org/showthread.php?p=942582#post942582

zambelli

18th August 2007, 19:48

Thanks, Andreas. I'll try this out sometime this week. If I find encoding errors again, I'll upload some samples.
Andreas: great work! I did a full 90-minute NTSC DV encode and found no obvious errors like when I tried it in 0.1.7.

Also, I re-ran some of the DV quality tests I set up in this thread (http://forum.doom9.org/showthread.php?p=908769) with the 0.2.0 build. Here are the scores:

|-----------------------------------------------------------------------------------------|
| | Ovrl PSNR | Min PSNR | Max PSNR | Notes |
| NTSC Interlaced: |
|-----------------------------------------------------------------------------------------|
| Cedocida v0.1.6 (VfW) | 49.3498 | 43.7395 | 55.0623 | YUY2, normal quality |
| Cedocida v0.2.0 (VfW) | 48.5104 | 42.1597 | 54.9393 | YUY2, normal quality |
| | | | | |
| NTSC Progressive: |
|-----------------------------------------------------------------------------------------|
| Cedocida v0.1.6 (VfW) | 49.8454 | 45.3902 | 53.4945 | YUY2, normal quality |
| Cedocida v0.2.0 (VfW) | 50.2529 | 46.5008 | 53.4981 | YUY2, normal quality |
| | | | | |
| PAL Interlaced: |
|-----------------------------------------------------------------------------------------|
| Cedocida v0.1.6 (VfW) | 46.1306 | 40.3453 | 54.7958 | YUY2, normal quality |
| Cedocida v0.2.0 (VfW) | 44.9788 | 39.3087 | 54.7926 | YUY2, normal quality |
| | | | | |
| PAL Progressive: |
|-----------------------------------------------------------------------------------------|
| Cedocida v0.1.6 (VfW) | 46.8568 | 41.5192 | 54.0270 | YUY2, normal quality |
| Cedocida v0.2.0 (VfW) | 47.8274 | 42.9472 | 54.0785 | YUY2, normal quality |
------------------------------------------------------------------------------------------

So the interlaced scores went down a bit, but the progressive scores improved. Overall, if you consider the MainConcept and Microsoft DV scores in the thread I linked, Cedocida is still the best DV codec with the least generational quality loss. Awesome work! :thanks:

cedocida

18th August 2007, 21:51

What about my other suggestion: internal low pass prefilter (as an option) ?
http://forum.doom9.org/showthread.php?p=942582#post942582

I guess you mean this:
Cedocida does not filter PAL chroma YUY2 source vertically before conversion to YUV420.
So, colors of some thin lines are simple discarded, see pictures at russian forum.
What is currently done in the cedocida encoder in case of YUY2 is exactly that, what is described in the specification (for color components: take the even (or odd) lines and drop the other ones). Of course this is not optimal as you get aliasing in case of sources which are not proper bandlimited. But if your source is originated from a DV cam, it allows you to not degrade the video in case of recompression (neglecting quantization).
Another point is interlacing. In this case, vertical linear filtering on fields is not appropriate, because there is allready aliasing, which is OK here. Low pass filtering can have some negative effects then.

rcubed

19th August 2007, 00:20

Hi,
I've been trying versions 0.1.7 and 0.2.0. With the defaults set, 0.1.7 will supply AVI captured with the Panasonic codec to Avisynth filters just fine as YV12. With the defaults set in 0.2.0 the process abends with no error indications. It appears to have something to do with the MPEG options being selected. Any ideas where to start? I'm interested in replacing the Panasonic codec. 0.2.0 seemed to capture ok if the default input/output options.

If YV12 is selected in the Decoder part of the Configure, and anything other than DV is selected for YV12 Chroma Sampling it will abend. If you try to use Gspot on the AVS script, Gspot will terminate if you attempt to render the stream. For the same options selected in 0.1.7 Gspot will open and render the AVS input. The same if the AVS is used as an input to VirtualDUb Mod.

The AVS script is as follows:

AVISource("H:\Test.avi")
KillAudio()
Info()

Thanks,

rcubed

juhok

19th August 2007, 11:58

I think I have same problem as rcubed (altho I don't know what 'abend' means).

After updating from 0.1.7 to 0.2.0 (WinXP 32bit) and opening DV-AVIs (PAL) thru AVISource(""), VirtualDub and MPC just exit without any message when trying to open the avs script.

When opening the same AVIs directly with said programs, there's no problem. Only when using Avisynth (version 2.57 build Dec 31 2006[13:16:28).

I tried re-installing the codec with default settings but the problem still remains.