Processors

Processors forum

State
Locked Locked
Replies 30 replies
Subscribers 100 subscribers
Views 1890 views
Users 0 members are here

Support feedback

Options

Options

Related

DaVinci TMS320DM365 - want to modify the yuv pixels before MPEG4 encoding. Encoder doesn't complain and encodes with mods, but decoder indicates bad data

Rod Burchell

Prodigy 230 points

Other Parts Discussed in Thread: CCSTUDIO

Hi,

In order to add some graphics to a video, we are modifying the yuv image prior to encoding it.

Here’s what we’re using during encoding to try and avoid encoding something out of range. This is more aggressive than the clipping specs we found. But, we still are seeing the decoder "crash" on the decode and send back a bad/corrupt data error on occasion.

//***todo - temp - boundary checks and limiting !!!***this is a temp hack***!!!

// Very frequent failures decoding when using the 16/235 and 16/240 limits,

// needs further looking at.

//Best info at the time: the MPEG 2 and hence MPEG 4 clipping specs are:

// for Y values must be between 16 and 235

// for U,V values must be between 16 and 240

//

//From http://en.wikipedia.org/wiki/YCbCr

if(yuv_pixel.components.y0 < 36)

yuv_pixel.components.y0 = 36;

if(yuv_pixel.components.y0 > 215)

yuv_pixel.components.y0 = 215;

if(yuv_pixel.components.y1 < 36)

yuv_pixel.components.y1 = 36;

if(yuv_pixel.components.y1 > 215)

yuv_pixel.components.y1 = 215;

if(yuv_pixel.components.u < 36)

yuv_pixel.components.u = 36;

if(yuv_pixel.components.u > 220)

yuv_pixel.components.u = 220;

if(yuv_pixel.components.v < 36)

yuv_pixel.components.v = 36;

if(yuv_pixel.components.v > 220)

yuv_pixel.components.v = 220;

…can’t seen to find information on what you can or can’t do during encoding that would result in the decoder detecting the encoded data as bad/corrupt. Can you shed any light on this?

Regards,

Rod Burchell

rod.burchell@raytek.com

831-458-3985

over 14 years ago

0 Marcus Cooksey over 14 years ago

TI__Genius 9510 points

Your question is being directed to experts in this area, but could you please provide logs which could help reveal why the decoder crashes?

0 Rod Burchell over 14 years ago in reply to Marcus Cooksey

Prodigy 230 points

Hi Marcus,

Not sure if this fits the bill for the "logs" that you'd like - let me know more specifically what to log and I can try and gather that information.

Here's a screen shot of when "process" returns the error.

8358.decode error.docx

The extended error is 0x00008802.

Here's a file that was encoded with the TI encoder. It fails with the TI decoder after a couple of seconds, but plays back fully with Windows Media player.

7411.IR004585.AVI

Regards,

Rod

0 Rod Burchell over 14 years ago in reply to Rod Burchell

Prodigy 230 points

Also, here's some date information at the end of ividdec2.h

/*

* @(#) ti.xdais.dm; 1, 0, 5,194; 3-13-2009 12:27:50; /db/wtree/library/trees/dais/dais-o05x/src/

*/

Regards,

Rod

0 Marcus Cooksey over 14 years ago in reply to Rod Burchell

TI__Genius 9510 points

We can start with what you've provided and translate what the 0x00008802 error means.

0 Rod Burchell over 14 years ago in reply to Marcus Cooksey

Prodigy 230 points

Hi Marcus,

Thanks for your help with this!

...any idea when we might get a response?

Sorry, ....don't like to have to push, but this is one of our few remaining outstanding items.

If there's anything I can provide to help speed up things, please let me know.

Regards,

Rod

0 Marcus Cooksey over 14 years ago in reply to Rod Burchell

TI__Genius 9510 points

Rod - I'm still waiting to hear back from the experts. I hope for a response soon.

0 Prasad RSV50057 over 14 years ago in reply to Marcus Cooksey

TI__Prodigy 695 points

Hi Rod,

Can you please let us know version numbers of the MPEG-4 encoder and decoder components you are using in your application?

Regards, RSV

0 Rod Burchell over 14 years ago in reply to Prasad RSV50057

Prodigy 230 points

Hi RSV,

I couldn't discern the version from what we have and I don't have the original download package. I've sent the files we're using to Marcus and asked him to forward them to you. (...couldn't see a way to attach files on the forum)

Hopefully, you'll be able to tell from those.

Is there a way I can tell by looking at a header file(s)?

Regards,

Rod

0 Prasad RSV50057 over 14 years ago in reply to Rod Burchell

TI__Prodigy 695 points

Hi Rod,

There will be release notes corresponding to respective encoder and decoder releases under your installed dirctories 210_V_MPEG4_D_01_00 and 210_V_MPEG4_E_01_00 (it will be available in one of the sub-directories). You can either me send those release notes through mail or send me the "Build ID" values mentioned in the correspoding release notes.

Regards, RSV

0 Rod Burchell over 14 years ago in reply to Prasad RSV50057

Prodigy 230 points

Hi RSV,

for the MPEG4 Restricted Simple Profile Decoder Version 01.00.00:

Build ID: 01.00.00.09

for the MPEG4 Simple Profile Encoder Version 01.00.00

Build ID: 01.00.00.12

Regards,

Rod

0 Prasad RSV50057 over 14 years ago in reply to Rod Burchell

TI__Prodigy 695 points

Hi Rod,

These are bit older versions of codecs. I am not sure if it is possible for you try with latest versions of encoder (01.10.00.03) and decoder (01.10.00.01).

We tried decoding the stream you provided with latest decoder and found it comes out at 91st frame with error. We are debugging to locate the potential issue with decoder.

Regards, RSV

0 Rod Burchell over 14 years ago in reply to Prasad RSV50057

Prodigy 230 points

Hi RSV,

Thanks for the update! We appreciate your help with this!!!

Any idea how long the debugging process may take?

Regards,

Rod

0 Rod Burchell over 14 years ago in reply to Rod Burchell

Prodigy 230 points

Hi All,

Anything at all that we can do to help speed up this process? Please let us know - two weeks seems like a long time to not have heard anything, but I really don't know how difficult this is to solve...or maybe you have alot on your plate already. Any information as to when there might be a solution would be greatly appreciated. ...sorry to have to bug you again, but there are a number of engineers interested in getting this issue solved asap.

Regards,

Rod

0 Prasad RSV50057 over 14 years ago in reply to Rod Burchell

TI__Prodigy 695 points

Hi Rod,

Sorry for the delay in responding on this subject.

Our team is working on this but couldn't locate exact problem location in decoder/encoder. I will get back to you with a date for fix by end of this week positively.

Regards, RSV

0 RSV Prasad over 14 years ago in reply to Rod Burchell

Intellectual 385 points

Hi Rod,

Is it possible for you try with latest encoder version 01.10.00.03 and let us know the result?

This will make sure we do not have any issue with encoder.

Regards, RSV

0 Rod Burchell over 14 years ago in reply to RSV Prasad

Prodigy 230 points

Hi RSV,

I took a brief look yesterday at updating the encoder. I didn't integrate the original encoder - so, when I looked to see if I could figure out how to do the update, my conclusion was that updating isn't all that easy. (i.e. files and folders mentioned in the instructions do not match what is unzipped).

Is there someone that can call me and walk me through what I need to do to update to the latest encoder?

Regards,

Rod

831-458-3985 hours are 9:30 - 5:00pm Pacific Standard Time

0 Rod Burchell over 14 years ago in reply to Rod Burchell

Prodigy 230 points

Hi RSV,

fyi - my PC is a Windows XP system, we're using the DM365 with a uCOSII rtos. This was originally done back in late 2009. The TI codecs are "wrapped" for use in our firmware. I discoverd the following tools being used when this was done: XDC 3.10.02, CGTools 4.1.3 and FC 2.23.01.

I noticed the available TI's DVSDK for Linux systems, but unfortunately, we're not using Linux. I do have a Linux PC available to me if it would be best to do this on a Linux PC then port the functions/libs over to Windows.

Regards,

Rod

0 RSV Prasad over 14 years ago in reply to Rod Burchell

Intellectual 385 points

Hi Rod,

It looks to me that it will be difficult for you to integrate new encoder. But we can have a conf call to explain on integration procedure in case if you want to do that.

1. Please provide us the encoder parameters being used to encode the stream?

2. Also is it possible for you to perform one experiment with decoder and let us know the results: Ignore the error (0x8802) returned by decoder try to continue the decoding with next frame

Regards, RSV

0 Rod Burchell over 14 years ago in reply to RSV Prasad

Prodigy 230 points

Hi RSV,

Regarding "conf call" - yes, I'd appreciate that.

Regarding 1. Here's the function used to set up encoding - I believe it has all the encoder set up parameters.

7824.enc_start.txt

Regarding 2. I ignored the decoding error and the decoder was able to successfully decode subsequent frames.

Regards,

Rod

0 Rod Burchell over 14 years ago in reply to Rod Burchell

Prodigy 230 points

Hi RSV,

This occurs before the "start" that I just sent you. This init routing is called one time at the begining of the app.

7026.enc_init.txt

Regards,

Rod

0 RSV Prasad over 14 years ago in reply to Rod Burchell

Intellectual 385 points

Hi Rod,

We will review the parameters and revert back to you.

Regards, RSV

0 RSV Prasad over 14 years ago in reply to RSV Prasad

Intellectual 385 points

Hi Rod,

Can you try with following for encoder and let us know your observations:

In "MpegCodec_EncodeInit()"

	params.videncParams.encodingPreset = 3

and

In "MpegCodec_EncodeStart()"

dynParams.rcQMin = 4;

dynParams.intraAlgo = 0

dynParams.skipMBAlgo =0

dynParams.unrestrictedMV= 0

     dynParams.meAlgo        = 0;

Regards, RSV

0 Rod Burchell over 14 years ago in reply to RSV Prasad

Prodigy 230 points

Hi Prasad,

...sorry for the long delay in getting back to you.

I used the settings that you recommended for the encoder. After doing so, I no longer observed any decoding failure.

Is there a suspected problem with the high quality encoding?

Regards,

Rod

0 Rod Burchell over 14 years ago in reply to Rod Burchell

Prodigy 230 points

Hi RSV,

If this is a high quality encoding problem, does the newer version of encoder have a fix for this? Either way, we'd still like to have a conference call and walk through the steps for creating a wrapped codec using the current encoder/decoder. Obvously, we'd like to get as good a quality of video as we can.

Regards,

Rod

0 Rod Burchell over 14 years ago in reply to Rod Burchell

Prodigy 230 points

Hi RSV,

FYI - I did some side by side comparisons and was able to see the degraded quality when moving from high quality to high performance.

Today I tried using SMPlayer and single stepped through a video (with some "bad" frames) frame by frame...best I can tell it decodes every frame.

Have you been able to discern if this is a encoder, decoder or both issue?

Please let us know when we can have a conference call.

Regards,

Rod

0 RSV Prasad over 14 years ago in reply to Rod Burchell

Intellectual 385 points

Hi Rod,

We can have call on Friday (22nd Jul) at 5Pm IST. Please suggest if that is fine with you.

We will provide our observation on issue in the call.

Regards, RSV

0 Rod Burchell over 14 years ago in reply to RSV Prasad

Prodigy 230 points

Hi RSV,

Please confirm that your time "IST" is Indian Standard Time. If so, we'd like to meet next week.

If its Indian Standard Time, you are 12:30 hours ahead of us.

How about Tuesday, Wednesday or Friday 7:30am or 8:00am Indian Standard Time?

Please email me and I will send you the conference call in number.

Regards,

Rod

rsbdesigns@sbcglobal.net

0 RSV Prasad over 14 years ago in reply to Rod Burchell

Intellectual 385 points

Hi Rod,

We can have the call on Tuesday - 26th Jul @ 8Am IST. Please send is the number.

Regards, RSV

0 Rod Burchell over 14 years ago in reply to RSV Prasad

Prodigy 230 points

Hi RSV,

We do not wish to post our conference call phone number and access code on this forum.

Please send me an email directly and I will email back to you our conference phone calling information.

Regards,

Rod

email to:

rsbdesigns@sbcglobal.net

rod.burchell@raytek.com

0 Rod Burchell over 14 years ago in reply to Rod Burchell

Prodigy 230 points

Hi RSV,

For Tuesday, we'd like to discuss the following items:

1 - What is TI's assessment of our encode/decode problem and what are the possible workarounds without upgrading to the latest encoder/decoder?

2 - Steps to take the current TI codecs and make wrapped codecs from them. Here's the basic outline of what was done 2+ years ago:

   - Install CCStudio, XCD, CGtools and Framework Components
   - modify the mp4encTestApp.mak and mp4vdecTestApp.mak
   - build the codec example code using gmake to get wrapped library projects
   - open each of these generated projects in CCS and build them
   - copy all the resultant libraries into the Application

..look forward to speaking with you.

Regards,

Rod