VPSS Front-end

Hi abtreelph,

Are you using any of the BIOS PSP? If yes, please mention the BIOS PSP version and also other tool versions being used.

With respect to H/W setup, please mention the HW platform which you are using. Also mention whether it is a custom board or the EVM.

Thanks and Regards,

Sandeep K

BIOS PSP 1.0.2

Customer board with DM6435

Everything works.  Images are acquired and converted properly to yuv422.  Changes can be made in the parameters to achieve a good quailty image.  However, we are guessing on the paramter changes.  The question is how are the gain settings applied to each channel value.  In other words, a value of 0x30 in wbGain[0] results in what operation being applied to channel 0 pixels?

    prevParams.whiteBalanceParam.wbDgain = 180;
    prevParams.whiteBalanceParam.wbGain[0] = 0x30;
    prevParams.whiteBalanceParam.wbGain[1] = 0x75;
    prevParams.whiteBalanceParam.wbGain[2] = 0x30;
    prevParams.whiteBalanceParam.wbGain[3] = 0x30;

Hi abtreelph,

abtreelph said:

    prevParams.whiteBalanceParam.wbGain[0] = 0x40;
    prevParams.whiteBalanceParam.wbGain[1] = 0x50;
    prevParams.whiteBalanceParam.wbGain[2] = 0x40;
    prevParams.whiteBalanceParam.wbGain[3] = 0x50;

These are the wbgain coefficients. This has been explained in section 4.3.3.7 o the http://www.ti.com/lit/ug/spru977d/spru977d.pdf document.

abtreelph said:

    prevParams.whiteBalanceParam.position[0][0] = 0;
    prevParams.whiteBalanceParam.position[0][1] = 1;
    prevParams.whiteBalanceParam.position[0][2] = 0;
    prevParams.whiteBalanceParam.position[0][3] = 1;

    prevParams.whiteBalanceParam.position[1][0] = 2;
    prevParams.whiteBalanceParam.position[1][1] = 3;
    prevParams.whiteBalanceParam.position[1][2] = 2;
    prevParams.whiteBalanceParam.position[1][3] = 3;

    prevParams.whiteBalanceParam.position[2][0] = 0;
    prevParams.whiteBalanceParam.position[2][1] = 1;
    prevParams.whiteBalanceParam.position[2][2] = 0;
    prevParams.whiteBalanceParam.position[2][3] = 1;

    prevParams.whiteBalanceParam.position[3][0] = 2;
    prevParams.whiteBalanceParam.position[3][1] = 3;
    prevParams.whiteBalanceParam.position[3][2] = 2;
    prevParams.whiteBalanceParam.position[3][3] = 3;

These parameter are corresponds to the register WBSEL (white balance coefficient selection register). Each pixel of each line can be configured with a coeffient and those coeffients are selected as 0,1,2 and 3 corresponds to COEF0, COEF1, COEF2 and COEF3 (i.e wbgain[0], wbgain[1], wbgain[2] and wbgain[3]).

More information on the WBSEL, WBGAIN register can be found in the section 6.2.16 and 6.2.15 respectively of the document http://www.ti.com/lit/ug/spru977d/spru977d.pdf

Thanks and Regards,

Sandeep K

Thanks, Sandeep.

Can you verify the multiplication of the coeeficient by the pixel value.  Is it straigt multipication or is there any scalling applied.  For example, if the following are inputs:

coeefficient = 0x40

pixel value = 0x80

What is the output value for 0x40 x 0x80 =

I am assuming the result is divided by 256.

Thanks and Regards,

Ralph

Hi ,

abtreelph said:

Can you verify the multiplication of the coeeficient by the pixel value

This has been explained in section 4.3.3.7 o the http://www.ti.com/lit/ug/spru977d/spru977d.pdf document. The statement says "In the white balance gain adjuster (WBGAIN), the raw data is multiplied by a selected gain corresponding to the color of the processed pixel"

For more information on this, you need to post the query on DM64xx Forum http://e2e.ti.com/support/dsp/davinci_digital_media_processors/f/99.aspx .

Thanks and Regards,

Sandeep K

Sandeep,

you are correct.  However, the document does not explain the overflow.  Two eight bit values are being multiplied and stored in an 8 bit result.  Please verify that the result is divided by 256 for proper scaling prior to storing the result.

Thankyou