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.

RTOS/TDA3XEVM: Tone mapping in one pass WDR (OV10640)

Part Number: TDA3XEVM

Tool/software: TI-RTOS

Hi,

I'm trying to understand the flow of data in one pass WDR mode with OV10640 MIPI sensor. As per my understanding

 in 1-pass WDR:

  • The sensor outputs a 12 bit PWL companded stream.
  • The data is decompanded to 20 bit and then tone mapped down to 12 bit as per this link.
  • NSF3 is disabled.
  • The h3a is performed after global tone mapping according to this link.

Since the decompanded data is 20 bit, is it being tone mapped down to 12 bit again in the IPIPEIF?

Is GLBCE used for performing the global tone mapping or is it the global tone mapping (PWL companding) performed by OV10640?

Best regards,

Srikanth Varanasi.

  • The sensor applies a crude global tone map which is just PWL companding. IPIPEIF reverses it to recover original 20bit image and applies a smooth 20->12 global tone map. Later GLBCE comes in to apply a locally adaptive tone map (12 ->12). 

    Please let us know if you have further questions.

  • Hi Mayank,

    Thanks for the quick reply.

    I am currently referring to TDA3x TRM available at "www.ti.com/.../spruie7c.pdf".  Could you please point out the block that actually performs the 20->12 global tone mapping in the figure 7-79 of section 7.3.3.4.2.

    In the Vision SDK, We have found that in OV10640's DCC file for one pass WDR IPIPEIF config(OV10640_ipipeif_split_dcc.xml) there are settings used for decompanding 12 bit PWL to 20 bit and also for tone mapping back to 12 bit. Is my assumption that H3A data in one pass WDR is generated using this 12->20->12 bit tone mapped data, which is not GLBCE tone mapped correct?

    Then in accordance with the above assumption, H3A data needs to be just inverse tone mapped back to 20 bit(for the 20->12 tone mapping performed in IPIPEIF) before doing any post processing, rather than doing both inverse tone mapping and decompanding. Is my understanding correct?

    Thanks again.

    Best regards,

    Srikanth Varanasi.

  • Hi Mayank,

    Could you please comment on the above query?

    Best regards,

    Srikanth Varanasi.

  • In the case of 1-pass WDR, "Decompand" block is doing the GTM. Basically it is a table lookup operation where 12-bit input is mapped to a 12-bit output. The logic for mapping is:

    Step 1 - Expand from 12->20 reversing the PWL settings which sensor used for companding

    Step 2- the 20 bit image is compressed to 12-bit using a smooth gamma curve i.e.

    y = ((x/(1048575) ^ gamma) * 4095

    Step 1 and Step 2 can be combined into a single LUT

  • Thank you Mayank