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.

TDA4VM: TDA4x NF tuning questions

Part Number: TDA4VM

Hi TI experts,

I have some questions about NF block concept and tuning.

1. 2 LUTs, Combined LUT and Division LUT are used for NF tuning. Could you help provide examples of these 2 LUTs?

2. I can find 'coeff.txt' in 'psdk_rtos_auto_j7_06_02_00_21\pdk\packages\ti\drv\vhwa\examples\vhwa_nf_test' folder. Could you help explain what it is used for?

3. Concept. According to the formula, I think the basic concept of NF is that 1-D Gauss filter and weighted with intensity difference, which is assumed follow Gauss distribution. Please correct me if I was wrong.

4. Sigma-s and Sigma-c. There are not registers for these 2 parameters. I think I can adjust either or both to generate LUTs then check output image, correct?

Best regards,

Zhishen

  • Hello Zhishen,

    There are two noise filters in VPAC.

    1. NSF4 inside VISS> WOrks on RAW (non YUV) images
    2. BNF - Outside VISS. Works on YUV images

    My understanding is that you are trying to use BNF. This is OK if you have a YUV coming in from CSI. However, if you are capturing a RAW image and processing it on VISS, I would suggest using NSF4 for noise reduction.

    Please confirm which NF you want to use and we can guide you accordingly.

    Regards,

    Mayank

  • Hi Mayank,

    Please refer to the snapshot below. I mean the Noise filter outside VISS which is in YUV domain.

    I plan to use both filters to achieve better de-noise performance. The data flow is raw input to VISS and process through NSF4 in raw domain, then go to Noise Filter in YUV domain.

    Best regards,

    Zhishen

  • Thanks for explaining. The folder psdk_rtos_auto_j7_06_02_00_21\pdk\packages\ti\drv\vhwa\examples\vhwa_nf_test is just a low level driver test code which is not used in any application. Currently supported vision apps demos (app_single_cam, app_multi_cam) do not enable NF. Can you please explain how are you running NF?

  • Hi Mayank,

    I didn't run NF yet because I haven't get our target module. Thanks to 'vhwa_nf_test' that I think there is possibility to use NF even without DCC. I plan to apply TI Extension Kernel ( tivxVpacNfBilateralNode) for NF running.

    I will come back to you if any problems about my idea and implementation.

    Thanks.

    Best regards,

    Zhishen

  • Thanks. Please let us know when you are ready.