• State
  • Locked Locked
  • Replies 3 replies
  • Subscribers 102 subscribers
  • Views 227 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.

Monochrome sensor thru VPFE

Eric Bryan
Expert 1045 points

I am working on a project with a monchrome CMOS sensor attached to a DM6435.  Much of the video material is based upon color sensors.  Two items have come up as particular questions:

1.  Document SPRU977A  section 4.3.1.2.3 pages 41-42 discuss the Fault Pixel Correction.  We are using a 12 bit monochrome CMOS imager.  The first descriptions of the correction mechanisms state that the pixel corrections are done with pixels “of the same color”.  Is the system capable of doing this correction for us or is it going to assume a Bayer pattern?  We just want to use the adjacent pixel.  Table 17 does not have this qualifier “of the same color” so it occurs to me that the system may not know the color pattern of the sensor and method 0 will really do what I want.

2.  The Preview Engine does dark frame subtraction which we want.  Unfortunately, it forces RGB blending and RGB to YCbCr conversion.  Are there settings that will allow us to nullify the effect of those stages or are we better off doing the subtraction with the processor?

  • 0
    TI__Mastermind 41665 points

    Eric Bryan said:
    1.  Document SPRU977A  section 4.3.1.2.3 pages 41-42 discuss the Fault Pixel Correction.  We are using a 12 bit monochrome CMOS imager.  The first descriptions of the correction mechanisms state that the pixel corrections are done with pixels “of the same color”.  Is the system capable of doing this correction for us or is it going to assume a Bayer pattern?  We just want to use the adjacent pixel.  Table 17 does not have this qualifier “of the same color” so it occurs to me that the system may not know the color pattern of the sensor and method 0 will really do what I want.

    I believe the statement that pixel corrections are done with pixels of the same color is just trying to imply that the next pixel that is needed for fault correction is not necessarily the adjacent pixel but rather the next closest pixel that is of the same color which you end up setting with the correction method bits. In your particular case of a monochrome sensor you would always be looking at the adjacent pixel so you would probably use a correction method of 0, 1, or 2.

    Eric Bryan said:
    2.  The Preview Engine does dark frame subtraction which we want.  Unfortunately, it forces RGB blending and RGB to YCbCr conversion.  Are there settings that will allow us to nullify the effect of those stages or are we better off doing the subtraction with the processor?

    Unfortunately as shown in figure 28 the RGB blending and RGB to YCbCr conversion cannot be disabled, however you do have control over the coefficients it uses.

  • 0
    Intellectual 390 points

    Hello Eric,

    I am interfacing a VGA color sensor with DM6446 and I am stuck at dark frame subtraction.

    The 'Understanding Preview engine document' shows a pointer variable void *dark_frame_addr of prev_params struct but its usage is not explained anywhere. How to write dark frame data to memory and pass its address to preview engine and the sequence of steps to be followed to achieve this is not clear to me.

    could you please explain how you did in your project.

    Thanks for your help.

    --

    Best Regards

    Arun

  • 0 in reply to Arun12185
    Expert 1045 points

    I haven't done it yet either!

     

    Eric