• State Resolved
  • Locked Locked
  • Replies 9 replies
  • Subscribers 102 subscribers
  • Views 260 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

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.

AM62A3: About the demosaic algorithm used by TDA4/AM62A

Ding Y
Prodigy 215 points
Part Number: AM62A3

Dear TI experts,

When we investigated the purple circle problem caused by enabling GLBCE, we found that if bilinear interpolation was used when demosaic was performed on bayer images (without considering the pixel value gradient), pseudo-color problems might occur near the high contrast of images, as shown in the following figure. What is the algorithm used by AM62A/TDA4 to do demosaic? Could the cause of the purple circle be related to this?

Best,

  • 0
    TI__Guru 52135 points

    Hi Ding,

    Do you have some sample images to share?

    You may find CFAI information in TDA4 TRM in the FCP section.
    It is gradient based interpolation.

  • 0 in reply to Gang Hua
    Prodigy 215 points

    pic.zip

    Hi Gang,

    Here is the picture, thanks.

  • 0 in reply to Ding Y
    TI__Guru 52115 points

    Hi Ding,

    What's the image resolution?

    Thanks,

    Jianzhong

  • 0 in reply to Jianzhong Xu
    Prodigy 215 points

    Hi Jianzhong,

    The image resolution is 1920 x 1080, and the bit depth of the raw image is 12.

    Best,

  • 0 in reply to Ding Y
    TI__Guru 52135 points

    Hi Ding,

    Could you please also share the sensor companding knee points?

  • 0 in reply to Gang Hua
    Prodigy 215 points

    kneepoints.txt
    Fullscreen
    1
    2
    3
    4
    5
    6
    7
    8
    9
    10
    11
    12
    13
    14
    15
    16
    17
    0 0
    1023 1023
    1279 2047
    1535 4095
    2303 16383
    2559 24575
    2943 49151
    3071 65535
    3327 131071
    3455 196607
    3519 262143
    3647 524287
    3775 1048575
    3903 2097151
    3999 4194303
    4095 16777215
    65535 16777215
    XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

    Hi Gang,

    Here is the sensor companding knee points.

    In addition, please tell me what other data you need before 18:00 on February 7, Beijing time. I will take my Spring Festival holiday later than this time.

    Best

  • 0 in reply to Ding Y
    TI__Guru 52135 points

    Hi Ding,

    Ding Y said:
    What is the algorithm used by AM62A/TDA4 to do demosaic? Could the cause of the purple circle be related to this?

    It looks like the issue is much more related to GLBCE than CFAI.

    The purple edge is present after GLBCE and before CFAI.

  • 0 in reply to Gang Hua
    Prodigy 215 points

    Hi Gang,

    Since this only happens when the bilinear interpolation de-mosaicing algorithm is used at very high contrast. I wonder if it's possible that it's the contrast enhancement by GLBCE that's causing CFAI to cause the purple edge problem. In other words, the disappearance of the purple edge problem when GLBCE is turned off is the result of a low local-contrast factor.

    Best

  • +1 in reply to Ding Y
    TI__Guru 52135 points

    Hi Ding,

    I find the purple edge present in GLBCE output before CFAI.
    You may refer to this raw image as an example.

    /cfs-file/__key/communityserver-discussions-components-files/791/pic_5F00_1920x1080_5F00_B.raw


    It sometimes shows up as a dark band or purple circle around bright spot lights after GLBCE.
    It seems GLBCE makes a ripple in those areas (even if the brightness goes up monotonically while we go toward light center without any ripple before GLBCE).

    I don't fully understand this GLBCE's behavior as it is a 3rd party IP.