• State Resolved
  • Locked Locked
  • Replies 5 replies
  • Subscribers 98 subscribers
  • Views 1218 views
  • Users 0 members are here
Support feedback
Options
Options
Related

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.

TDA3: Pink tint visible at saturation regions in One pass WDR mode(OV10640)

Navin Prashath
Intellectual 760 points

Part Number: TDA3

Hi,

When we enable one pass WDR mode with OV10640 with AEWB algorithm, we are observing pink tint at the saturation regions.

We observed that the pink tint is the result of Red and Blue  white balance gains applied to the IPIPE. So we applied white balance Red and Blue gains on the sensor instead of IPIPE.

While the tint has reduced, artifacts are still visible at the saturation regions. What could be the reason for this color artifact at the saturation region?

Is this the problem with sensor's WDR combine / tone map module or the GLBCE tonemapping in the IPIPEIF? Could this issue can be fixed by changing either sensor or GLBCE parameters?

P.S: We are observing the issue in two pass WDR also.

Regards,

Navinprashath.R.R

  • 0
    TI__Guru 59470 points

    Hi Navinprashath,

    If AWB gain is applied correctly in sensor, the pink tint is typically caused by saturation in sensor combine/WB gain.

    Could you please upload a raw image captured from sensor to show the issue?

    Best,

    Gang

  • 0 in reply to Gang Hua
    Intellectual 760 points

    Hi Gang Hua,

    Please find the raw image for the one pass WDR mode and linear mode (Resolution: 1280 x 720 BGGR 12bit).

    (P.S: The file extensions has been replaced with .123 for uploading)

    OV10640_1280x720_Linear.123OV10640_1280x720_One_pass_WDR.123

    Thanks in advance,

    Navinprashath.R.R

  • 0 in reply to Navin Prashath
    TI__Guru 59470 points
    Hi Navinprashath,

    Have AWB gains been applied to the one pass WDR raw image already?
    The pink area is completely saturated already in the raw image and TI ISP cannot recover from it.

    OV10640 has a dual-pixel structure and OV has some recommendations for applying AWB gains properly in sensor.
    You may check with OV if you have not done so.

    Best,
    Gang
  • 0 in reply to Gang Hua
    Intellectual 760 points
    Hi Gang,

    Yes, the white balance gains has already been applied in the sensor.
    The pink tint is visible(as a ring) at the transition from saturated region to unsaturated region.

    In case of pink tint, when white balance gains are applied in the IPIPE, we believe it's because of the companding(in sensor) and decompanding(in ISP).

    Regards,
    Navinprashath.R.R
  • 0 in reply to Navin Prashath
    TI__Guru 59470 points
    Hi Navinprashath,

    Yes, for WDR images, applying WB gains in IPIPE has limitations since the pixels have already been tone mapped to 12-bit LDR format.
    Therefore, it is better to apply WB gains in the sensor in WDR case.
    This issue is caused by the WDR sensor and cannot be fixed by adjusting any parameter in TDA3 ISP H/W.

    In your raw image, the light bulb inside the lamp is completely saturated (on all R/G/B pixels) and therefore we cannot tell if there is any WB issue.
    Between the light bulb and the lamp edge, the pixels seem change color slightly in space (however, this raw image is very limited in details and I am not sure how accurate this observation is).
    If you have not followed OV's instruction about compensating the WB difference for the dual-pixel structure, this type of color shift/tint may happen in the bright side as well as in the transition region, because the sensor exposure frames from the large pixels and the small pixels have different color tint after the WB in sensor.
    According to OV, if you do the WB compensation for the small pixel properly, the issue can be fixed (but TI does not have the details on this).

    Best,
    Gang