Multimedia Communications using Scalable Video Coding

Trackback : http://r2d2n3po.tistory.com/trackback/7

1. 2011/06/26 00:43

   비밀댓글입니다

2. jhamed 2011/06/26 00:47

   hi again...
   i've download this PPT's and follow your steps in JSVM tutorial 2.ppt
   But it does not work !!
   what is the problem in that? (as you said you found some mistakes in it)
   thanks a lot

3. hope 2011/07/07 10:24

   Hi jhamed,
   I am not sure in which step you met the problem. please let me know it.
   for the mistakes, I will update this slide as soon as possible.

4. hope 2011/07/11 13:28

   I updated the slide. you can download most recent one. if you have further questions, please let me know them in detail.

5. jhamed 2011/07/20 23:32

   Thanks a lot Hope :))
   i've solved my problem with encoding in SVC...
   but i've a stupid problem !
   i prefer to simulate packet lossy communication in MATLAB...
   i wrote a code but how can i use SVC Encoder output (test.264) as an input for my script code ?
   (of course in packet form)
   Thanks again...

   • hope 2011/07/27 11:18

     Sorry, I didn't do the experiment with MATLAB.
     But, I Know that MATLAB can read the file data(SVC-encoded file), which consists of multiple NAL packets. Then, you can decide if the error generator drops a certain NAL packet or not according to your own algorithm or traced file.
     In my case, I implemented my own code with C.
     Otherwise, you can use JVT's tool that I listed somewhere in this blog.
     
     Good luck!

6. jhamed 2011/12/11 01:41

   Hi again...
   
   Thanks for your helps again... ;))
   
   Again a problem...
   how can i extract Base and Enhancement layers from output of encoder (test.264) ?
   
   i know "BitStreamExtractorStatic" can be used for extracting sub-bitstreams
   with a lower spatio-temporal resolution and/or bit-rate from a
   global scalable (SVC) bit-stream. ...
   
   But in this way upper layers have lower layer information (bits) again !! (may be my fault ! )
   i want to have for example layer 1 information without information of layer 0 (base layer)....
   
   please help me hope...
   Thanks a lot...

7. hope 2011/12/13 14:49

   Hi,
   I implemented the code two years ago, but I don't have it now ( I have the modified version which is able to support 1400-Byte packets with separated layer).
   I wrote the code by modifying this code (please see this link: http://r2d2n3po.tistory.com/25).
   The only thing that you have to is store the classified NAL packets separately according to its layer ID.
   
   Thanks,

8. jhamed 2011/12/23 10:34

   That link is Unavalable !!! : ((
   
   please say more about the separation of NALs....
   (Which commands are useful for that ?!)

9. hope 2011/12/29 16:51

   Please use this link: http://r2d2n3po.tistory.com/25
   

10. Kirtikumar Patel 2012/01/10 15:52

    How to set path for image sequences ?
    Where to find decoder.cfg and how to use this .cfg files?

11. hope 2012/01/11 13:43

    Kirtikumar Patel// the config files are included in JSVM, and you can download the reference tool from the link listed in the slide.
    Once you open the config file (.cfg), you will find the 'input filename option' at the top of the lines. Set it with your own test sequence file name.
    

12. Kirtikumar Patel 2012/01/11 15:35

    Dear hope
    I downloaded the jsvm 9.19 . I have encoder.cfg,layer 1-7 but i dont have decoder.cfg ....
    I had downloaded all the sequences but where to give path for it ?
    And I also try to see it by yuvviewer.exe..but I cant see the image sequences. ..how to do this 3 things ??
    
    Please help.
    
    thanks

13. hope 2012/01/11 17:00

    Dear Kirtikumar Patel,
    1) You don't need any decoder.cfg file.
    Once you launched the decoder executable, you will see its usage as shown below.
    > Decoder.exe BitstreamFile.svc TargetYUVFile <some options>
    
    2) the YUV sequence file name should be listed in layer#.cfg file.
    Once you open the file, you will see the line: InputFile org_176x144_15.yuv # Input file
    You have to edit the file name.
    
    3) YUV file doesn't have any information about the characteristics of sequence file such as resolution, color space(YUV), subsampling(4:2:0), per-pixel bit depth (8-bit), and so on.
    So, you need to check those information first, and let the YUVplayer know those information.
    
    I hope my answer will work for your test.

14. ikram 2012/01/25 03:46

    Hi i'm working with jsvm 9 with MGS SNR scalability, my problem is what is the difference between -lqp and -rqp in >H264AVCEncoderLibTestStatic –pf main.cfg -lqp 0 30 -rqp 0 32 -lqp 1 24 -rqp 1 26
    i found in the software manuel
    -rqp (layer) (rQP)
    Specifies the basis quantization parameter rQP to be used for quantization of the layer specified by layer and the Lagrangian parameter MeQPLP used for motion estimation and mode decision of key pictures
    . Overwrites the parameters QP and MeQPLP of the corresponding layer configuration file.
    
    -lqp (layer) (lQP)
    Specifies the basis quantization parameter and the Lagrangian parameters for the layer given by layer, including MeQPLP. The option –lqp X lQP corresponds to the following list of options –rqp and –mqp:
    –rqp X lQP –mqp X 0 lQP –mqp X 1 lQP –mqp X 2 lQP –mqp X 3 lQP –mqp X 4 lQP –mqp X 5 lQP
    Note that –lqp X QP1 –rqp X QP2 sets MeQPLP = QP1.
    but in the exemple the author say
    The option “-lqp” is known from the examples for single-layer and spatial scalable coding (section 3.1 and 3.3). This option specifies the quantization parameter and the Lagrangian parameters for mode decision and motion estimation. The option “-rqp” specifies the quantization parameter. Thus, the previous value given by “-lqp” is overwritten. As a consequence, with option “-lqp” only the Lagrangian multipliers for motion estimation and mode decision are specified (as a QP equivalent), while the actual used quantization parameter is specified by the option “-rqp”. For rate-distortion efficient MGS coding, the values of MeQP (option “-lqp” in the example) that are used for determining the Lagrangian parameters should be set to smaller values than the actual quantization parameter (option “-rqp” in the example). The “optimal” difference depends on the number of MGS layers that should be encoded as well as on the sequence content. For SNR scalable coding with one MGS layer, a QP difference of 1 or 2 is reasonable in most cases.
    is it not contradictory?
    Please i would like to understand
    thanks