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.

DM644x VPFE Question

We have a nonstandard video format that we want to process in the DaVinci DSP.  We'd like to the use the VPFE to read in the data.  We have the pixel clock rate down to a rate that is supported by the TI DaVinci part, but we want to see if it is possible to use the VPFE to read in this data.  This is a CMOS image sensor and it allows a user to "window" out a region of interest to increase the effective frame rate.  We wish to window out a 98x98 frame at 600 Hz.  The quantization is at 10-bit per pixel (greyscale).  Is it possible to read this image through the VPFE into the DaVinci DM644x?  I'm looking at the CCDC and I see support for REC656, YCC and Raw (Bayer) pixel formats.  How would I input a greyscale image sensor?

 

  • In raw mode, whatever is at the CCD interface can be capture to SDRAM; howvere, if you need to do any further processing using the rest of the VPSS (preview, resizer....), the rest of our VPSS does not understand grayscale pixel format; therefore you may need to implement color conversion algorithm in software.

  • We may want to use the resizer to reshape the images and display them to an NTSC monitor.  Does this mean I'd have to convert the greyscale to a color image?

  • resizer needs YCbCr pixel images, hence is your grayscale image is of a different pixel format, then yes.

  • What is the exact format and size supported by VPFE's IPIPE and H3A blocks?

  • These blocks take in conventional bayer pattern data; the exact format is discussed in DM644x VPFE User Guide.

  • Juan, I want to follow up on this for my application - very similar requirements.  Another person more familiar with this has set up the previewer to accept raw data from a Micron 034 mono 10 bit image sensor and produce a color YCbCr output (this part works properly).  Here are a few lines from the previewer setup:

    PSP_previewerParams Params;

    Params.sizeParam.pixelSize = PSP_PREVIEWER_INWIDTH_10BIT;

    Params.outPixelOrderParam = PSP_PREVIEWER_PIXELORDER_YCBYCR;

    If I understand correctly, I cannot resize the mono sensor output directly?  My goal is to reduce the resolution by 1/2 in each dimension prior to performing algorithms such as edge or corner detection to reduce processor cycles.  From this thread it appears that I must perform the following sequence to result in a 1/4 resolution mono image:

    1) preview to get the YCbCr format

    2) resize to get the YCbCr in 1/4 resolution

    3) convert to grayscale (possibly using EDMA3 to transfer only the Y values???)

    This seems like a rather round-about way to get where I want to be, is there a more direct route?

    Thank you, Ed