1 Reply Latest reply on Jun 28, 2018 3:03 PM by Intel Corporation

    IR Y16 vs Y8 Brightness and "Noise"

    Leponzo

      When we capture the same scene with the same camera settings (gain, exposure, etc.) in infrared (without the emitter) Y8 and Y16, the Y16 image is (1) brighter and (2) has tiny square pixel noise across it. Why is this so? Isn't Y16 just a better (bit-depth wise) version of Y8?

       

      Y8 image:

      SET2_Y8_GAIN_16_EXPOSURE_35000_IMAGE_0.png

       

      Y16 image:

      SET2_Y16_GAIN_16_EXPOSURE_35000_IMAGE_0.png

       

      If you zoom into a square in Y16, you can see the noise:

      Pattern Noise.PNG

       

      This noise is not visible in Y8:

      Capture.PNG

       

      I also compared the two images in MATLAB using imshowpair, and it shows there are differences even though the scene is exactly the same:

      Diff.png

        • 1. Re: IR Y16 vs Y8 Brightness and "Noise"
          Intel Corporation
          This message was posted on behalf of Intel Corporation

          Hello Leponzo,
           

          I apologize for taking so long to respond to your worthy question. This thread got lost in the shuffle somehow.

          The Y16 format shows only unrectified images and is to be used only when calibrating the camera. Do not use Y16 format in your applications.

          Use Y8 format in your applications. The Y8 format provides rectified images. This means that the depth images captured by each depth sensor have been rotated to align with each other so that the rotation matrix in the extrinsics is the identity matrix.

          Regards,
          Jesus G.
          Intel Customer Support

          1 of 1 people found this helpful