[ADV7282A-M] More details of the I2P interpolation.

Hi,

I have a question about details of I2P (interlace to progressive) conversion of ADV7282A-M.

The expert answered on following thread as follows:
https://ez.analog.com/video/f/q-a/9323/adv7280-deinterlacing-question/7255#7255
>The deinterlacer in the ADV7280, ADV7280-M, ADV7282, ADV7282-M all work on a line doubling rather than on a frame buffering technique.
>This gives the advantage of reduced delay on the output. Also large external  memory is not needed.
>The line doubling technique uses a propriety analog devices algorithm.
>It interpolates between two lines (e.g. lines 1 and 3 on an odd frame) to generate the an additional line (e.g. line 2).
>A smoothing filter is also applied to reduce the effect of low angle noise (i.e. jaggies).

Question:
 Please tell me more details of the I2P interpolation.
 About your comment "It interpolates between two lines (e.g. lines 1 and 3 on an odd frame) to generate the an additional line (e.g. line 2). ",
 => Does it mean linear interpolation? (Does it mean that line2 is interpolated by linear?)
      Is it the way to generate line 2 from average of line 1 and line 3?

Thank you!
Best regards.
Tamu

  • +1
    •  Analog Employees 
    on Sep 24, 2021 12:00 PM

    Hi Tamu-San,

    Please find the below comment,

      Does it mean linear interpolation? (Does it mean that line2 is interpolated by linear?)
      Is it the way to generate line 2 from average of line 1 and line 3?

            YES, linear interpolation will generate an new scan line between two input scan lines, & You can get the more details of linear interpolation in the "video-demy5.pdf" .

                     outn = (inn–1 + inn+1) / 2

    Please refer below snap,

    Note: Deinterlacer block (I2P) is designed to have a minimum amount of latency between input and output (max latency is about 5 lines of video ~ 85 ms). This is important for automotive applications. Due to this it uses a line based deinterlacer algorithm to interpolate between lines to generate the missing lines e.g. in odd fields it interpolates between lines 1 and 3 to generate line 2.

    Thanks,

    Poornima

  • Hi,

    Thank you for your reply !

    Please let me ask an additional question for more details as below:
    About your answer:
    > YES, linear interpolation will generate an new scan line between two input scan lines,
    > & You can get the more details of linear interpolation in the "video-demy5.pdf" .
    > outn = (inn–1 + inn+1) / 2

    Please tell me about more details of above equation.
    What average values are calculated on the equation?
    Are they Luma data (Y) and Chroma data (cb/cr)?

    Thank you!
    Best regards.
    Tamu

  • 0
    •  Analog Employees 
    on Sep 30, 2021 9:55 AM in reply to Tamu

    Hi Tamu San,

    Are they Luma data (Y) and Chroma data (cb/cr) ?

        Yes, It will use chroma and luma without affecting the  vertical resolution.

         Please note Deinterlacing must be performed on component video signals such as RGB or YCbCr but composite color video signals such as NTSC or PAL cannot be de-interlaced directly to the presence of color subcarrier phase information, which would be meaningless after processing. These composite signals must be decoded into component color signals such as RGB or YCbCr Prior to de-interlacing.

    Two methods that we are using in the de-interlacing algorithms like one is video mode  & another one is film mode.Please refer below snap for your reference with respective subdivision,

     


    Video Mode: Intra-Field Processing

      This is the simplest method for generating additional scan lines using only information in
    the original field. Although there are two common techniques for implementing intra-field processing,scan line duplication and scan line interpolation(linear interpolation), the resulting vertical resolution is always limited by the content of the original field

    Note: By using linear interpolation there is an improvement in video quality of scaled images.
    When an output sample falls between two input samples (horizontally or vertically), the output sample is computed by linearly interpolating between the two input samples. However, scaling to images smaller than one-half of the original still results in deleted samples.

    Thanks,

    Poornima