This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

does the previewer of dm365 do the de-interlace work?

hi,I am using the dvsdk demos(dvsdk_demos_2_10_00_1x/dm365/encode)

as there is the previewer and the resizer device to configure before the capure device

I found follows

case 1: when the  previewer is configured in sole mode (not chain with the reszier,and the resizer is not opend and configured),the captured picture has been de-interlaced,

case 2: when the  resizer is configured in sole mode(not chain with the previewer,and the previewer is not opend and configured), ,the captured picture has not been de-interlaced;

case3: also,when i chain them in continous mode, the captured picture also has  been de-interlaced.

But what i see in the VPFECaptureDriverUsersGuide,it says:

' In continuous mode, with interlaced scan input, only the second field of the input data is used for
capture at Resizer and is scaled up by 2x to preserve the height. So in this mode, the video is de-interlaced at the driver and hence both fields are not available.'

what i think this sentence is not typically pointed to dm365 but dm6446?

My question is: can anyone explain it and the upper threee cases? and does the previewer of dm365 do the de-interlace work? or can i be further told the difference between them?

thanks a lot

  • Hello YUJIAN ZHAO

     

    DM365 does not have any special hardware for deinterlace. The "deinterlace" your talking about is only a skipping of a field and resizing the other field. So deinterlace on DM365 is done by the RESIZER, but you sould consider: Using deinterlace on DM365 means loosing half of vertical resolution, and half of tempral resolution.

     

    Regards Josef

  • Hi Josef

    thanks a lot for your reply

    I think what you say is the case 3 I listed in my first post, is it?

    what I consern most is the difference between case 1 ,case 2 and case 3

    why is it  when I only open the previewer device(without resizer), the picture has been de-interlaced?

     

    thanks a lot

  • Hi Yujian Zhao,

    Case 1: You dont setup the resizer driver but previewer uses the resizer hardware module as the output device. This way case 1 and case 3 are almost same.

    Case 2: In this mode, the resizer takes the data from the DDR and does the resize operation. It does not use only one field and just utilizes all the frame data. Hence you dont see "deinterlace" effect.

    Case 3: Same as case 1 and explained by Josef already.

    Regards,

    anshuman

    PS: Please mark this post as verified, if you think it has answered your questions. Thanks.

  • Anshuman Saxena said:

    Hi Yujian Zhao,

    Case 1: You dont setup the resizer driver but previewer uses the resizer hardware module as the output device. This way case 1 and case 3 are almost same.

    Case 2: In this mode, the resizer takes the data from the DDR and does the resize operation. It does not use only one field and just utilizes all the frame data. Hence you dont see "deinterlace" effect.

    Case 3: Same as case 1 and explained by Josef already.

    Regards,

    anshuman

    Hi  Anshuman Saxena,

    Thanks a lot for you reply!

    Sorry I had made a mistake!

    I  found that case 1 and case 3 are almost same as you said when i reviewed the kernel. I think the difference between them is you can apply your setting in the user's application in case 3.

    I refer to this page : http://e2e.ti.com/support/embedded/f/354/t/41298.aspx

    Don Darling said:

    On DM365 there should be a way to configure the IPIPE to help with de-interlacing.

    You start with something like a D1 UYVY image:

            720 width, 1440 pitch
    +------------------------------------+
    |                                    |
    |                                    |
    |                                    | 480
    |                                    |
    |                                    |
    |                                    |
     
    +------------------------------------+

    Then you set-up the IPIPE like you're doing a re-size, but when you set the input paramaters, you tell the IPIPE the input looks like this:

            720 width, 2880 pitch
    +------------------------------------------------------------------------+
    |                                                                        |
    |                                                                        | 240
    |                                                                        |
    +------------------------------------------------------------------------+

    Now, when you resize from 720x240 UYVY to 720x480 UYVY, it will resize the left 720x2 bytes to be a full D1 image, which throws away the second field of the interlaced image:

            720 width, 2880 pitch
    +------------------------------------------------------------------------+
    |                                     |                                  |
    |          FIELD 1                    |     FIELD 2                      | 240
    |                                     |                                  |
    +------------------------------------------------------------------------+

                ||
                \/

            720 width, 1440 pitch
    +------------------------------------+
    |                                    |
    |                                    |
    |         FIELD 1                    | 480
    |                                    |
    |                                    |
    |                                    |
     
    +------------------------------------+

    and I do some experiment to verify what Don Darling said, but I can only confirm the resizer does use only one field of the picture and does a 2x scale, and it is the same as what the <VPFE Capture Driver Users Guide> document said.

    what confused me is that how can i tell IPIPE the input picture's pitch is 2880? this is my question.

    I dump the registers of ISIF, IPIPE and Resizer,and find there are only registers configure the pixel number on the Vertical or Horizontal direction of input picture , or how many pixels are write to DDR and its offset(pitch) on the Vertical or Horizontal direction. I don't find any setting than configure out the picture's pitch is 2880 (or may be something else) to tell the IPIPE or the resizer to ommit the other field of the picture (or do i missed something?)


    thanks

  • by the way, i try it in the in continous mode, and the input is from a PAL camera

    thanks