[D455] Compare Some Problems

• MartyG

  • August 06, 2022 06:43

  Hi Nguyenduyduc14  Thanks very much for your questions.

   

  1. It is recommended to apply post-processing filters first and alignment afterwards.  Performing alignment first could cause negative effects such as aliasing (jagged lines).

   

  2.  The most common alignment method is aligning depth to color, though it is not wrong to align color to depth (which is done much more rarely by RealSense users) if it benefits your particular project.

   

  Depth to color alignment resizes the depth sensor's field of view (FOV) to match the FOV of the color sensor.  On the D435/ D435i camera models that have a smaller FOV size on the color sensor, this can cause the outer areas of depth information to be cut out of the aligned image as though a black border has been overlaid around the edges of the image.  On camera models such as D455 that have approximately the same depth and color FOV size though, the amount of information excluded from the aligned image is very small.

   

  With color to depth alignment, the color FOV size is resized to match the depth FOV size.  This causes the color to be stretched so that the aligned image fills the screen and does not have detail excluded on the outer edges.  On the D435 / D435i the effect of the image stretching to fill the screen is very significant.  On cameras with approximately the same FOV size such as D455, the impact of this effect on the aligned image is smaller as there is.less empty space at the borders to fill out.

   

  In conclusion, I would recommend using depth to color alignment with D455 as color to depth is a much more rarely used technique than depth to color and would not provide a clear benefit on D455 compared to the image stretching benefit of color to depth alignment on D435 / D435i.

  0

  Comment actions Permalink
• Nguyenduyduc14

  • August 06, 2022 07:19

  Thank you.

  I have a few questions about intrinsics. I checked intrinsics from depthFrame and colorFrame. They are different and they are different from focal length, principal point of Camera calibration of IntelRealSenseViewer.

  If I apply postprocessing then aligan depth to color, which intrinsics should I use? 

  How do I get pointcloud quality as same as when I get from intelRealSenseViewer?

   

  0

  Comment actions Permalink
• MartyG

  • August 06, 2022 07:49

  You should use the intrinsics of the stream that you are aligning to.  So if aligning depth to color, use the color intrinsics.

   

  The RealSense Viewer applies a range of depth colorization / color scheme settings and post-processing filters by default.  So in addition to programming post-processing filters into your scripting, you should program colorization settings too in order to achieve an image that is closer to the ones produced by the Viewer..

   

  The Python resources at the link below provide a good illustration of the color settings that are customizable.

  https://github.com/IntelRealSense/librealsense/issues/7767#issuecomment-726704671

  0

  Comment actions Permalink
• Nguyenduyduc14

  • August 06, 2022 08:19

  Thank you so much.

  0

  Comment actions Permalink

Please sign in to leave a comment.