Generating a depth map from a point cloud using Python

Question

I am working on a project where I need to generate depth images from a point cloud. While I have come across examples in the literature that explains how to create a point cloud from a depth map, I'm specifically looking for guidance on the reverse process.

Can someone please provide insights or suggestions on how to create a depth map from a point cloud with Python programming language? Any relevant information, techniques, or resources would be greatly appreciated.

This question is too broad for Stack Overflow. "I want to do X" is not a question. Please ask a _specific_ question. Having attempted to solve your problem yourself usually helps in formulating a good question. See [what's on-topic here](/help/on-topic), [ask], the [question checklist](//meta.stackoverflow.com/q/260648/843953), and how to provide a [mre]. — Pranav Hosangadi, Nov 29 '22 at 17:47
While I understand the concern raised by @PranavHosangadi, My question is not more general than other questions asked her. Please see https://stackoverflow.com/questions/64094463/how-can-i-generate-a-point-cloud-from-a-depth-image-and-rgb-image — Valentyn, Nov 29 '22 at 18:01
If you want to do the reverse maybe you can try projection simply? I had projection related answer: https://stackoverflow.com/questions/43644656/converting-3d-point-cloud-to-a-2d-gridmap-on-python/73913607#73913607 — Shaig Hamzaliyev, Nov 30 '22 at 08:30

score 0 · Answer 1 · answered Jun 15 '23 at 13:21

First of all, you need the intrinsic matrix of the camera. You can find fx, fy, cx, and cy values from the camera intrinsic matrix.

Example intrinsic matrix:

    # [ Fx, 0,  Cx,  Fx*-Tx ]
    # [ 0,  Fy, Cy,  0      ]
    # [ 0,  0,  1,   0      ]

Here is an example of creating a point cloud from a depth map.

    # read the depth image here and define intrinsic values

    fx = FX_DEPTH
    fy = FY_DEPTH
    cx = CX_DEPTH
    cy = CY_DEPTH

    x = (u - cx) / fx
    y = (v - cy) / fy
    z = image[v, u]
    x *= z
    y *= z
    z = z
    print(f"x: {x} y: {y} z: {z}")

But you are looking for a Python example of the reverse process. Here is the code of the reverse process:

    # read a point from the point cloud and define intrinsic values

    fx = FX_DEPTH
    fy = FY_DEPTH
    cx = CX_DEPTH
    cy = CY_DEPTH

    u = (x / z) * fx + cx
    v = (y / z) * fy + cy
    w, h = image.shape
    image[v, u] = z
    print(f"Position: {u} {v} Value:{image[u, v]}")

You can create the depth map if you iterate this code piece for all points.

Generating a depth map from a point cloud using Python

1 Answers1