How do I directly covert a depth image to 3-D point cloud?

Hi,

I could use pcfromkinect() function to convert depth data from kinect device into 3-D cloud points.

But how could I convert the depth matrix that had already extracted from kinect into 3-D cloud points ?

I mean, pcfromkinect() seems only works from device to 3d points cloud, not matrix to cloud.

[jenny jenny]

did you get the answer, because i have the same problem

[Simon Cool]

Same problem!

Answer 1

[Marten J]

I had the same problem. I could solve it through a double for-loop... but it's not the most efficient way:

if true
A = double(depthImage);
n=0;
[s1 s2] = size(A);
for i = 1:s1
  for j = 1:s2
      n = n+1;
      Punkte(n,1) = i;
      Punkte(n,2) = j;
      Punkte(n,3) = A(i,j);     
  end
end
end

so I calculated the depthImage in the Pointcloud matrix [x-vector, y-vector, z-vector] but it takes some time.

Also only the Z-Value is in the metric system... the x and y- values stay in the Pixel system... to transform then in the Objekt-coordinate system you need the calibration and the Collinearity equations.

Answer 2

[Preetham Manjunatha]

This link can help to convert RGB-D images to point cloud, provided the camera intrinsic parameters.

Answer 3

[Birju Patel]

pcfromdepth has been added to Computer Vision Toolbox in R2022b:

https://www.mathworks.com/help/vision/ref/pcfromdepth.html