support TSDF deintegration for a previously integrated RGBD image

[iammarvelous]

Type

• [ ] Bug fix (non-breaking change which fixes an issue): Fixes #
• [x] New feature (non-breaking change which adds functionality). Resolves #2613
• [ ] Breaking change (fix or feature that would cause existing functionality to not work as expected) Resolves #

Motivation and Context

It's common to deintegrate a previously integrated RGBD image in practice. Having the function with such flag will be helpful for downstream applications.

Checklist:

• [x] I have run python util/check_style.py --apply to apply Open3D code style to my code.
• [x] This PR changes Open3D behavior or adds new functionality.
  • [x] Both C++ (Doxygen) and Python (Sphinx / Google style) documentation is updated accordingly.
  • [x] I have added or updated C++ and / or Python unit tests OR included test results (e.g. screenshots or numbers) here.
• [x] I will follow up and update the code if CI fails.
• [x] For fork PRs, I have selected Allow edits from maintainers.

Description

Mathematically, the deintegration can be written as: tsdf_new = (tsdf_old * w_old - tsdf_image) / (w_old - 1).

The PR was tested by following the example to integrate RGBD images and deintegrate each frame afterwards.

import open3d as o3d
import numpy as np

import os, sys

pyexample_path = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
sys.path.append(pyexample_path)

from open3d_example import read_trajectory

if __name__ == "__main__":
    rgbd_data = o3d.data.SampleRedwoodRGBDImages()
    camera_poses = read_trajectory(rgbd_data.odometry_log_path)
    camera_intrinsics = o3d.camera.PinholeCameraIntrinsic(
        o3d.camera.PinholeCameraIntrinsicParameters.PrimeSenseDefault)
    volume = o3d.pipelines.integration.UniformTSDFVolume(
        length=4.0,
        resolution=512,
        sdf_trunc=0.04,
        color_type=o3d.pipelines.integration.TSDFVolumeColorType.RGB8,
    )

    for i in range(len(camera_poses)):
        print("Integrate {:d}-th image into the volume.".format(i))
        color = o3d.io.read_image(rgbd_data.color_paths[i])
        depth = o3d.io.read_image(rgbd_data.depth_paths[i])

        rgbd = o3d.geometry.RGBDImage.create_from_color_and_depth(
            color, depth, depth_trunc=4.0, convert_rgb_to_intensity=False)
        volume.integrate(
            rgbd,
            camera_intrinsics,
            np.linalg.inv(camera_poses[i].pose),
        )

    print("Extract triangle mesh")
    mesh = volume.extract_triangle_mesh()
    mesh.compute_vertex_normals()
    o3d.visualization.draw_geometries([mesh])


    print("Extract voxel-aligned debugging point cloud")
    voxel_pcd = volume.extract_voxel_point_cloud()
    o3d.visualization.draw_geometries([voxel_pcd])

    print("Extract voxel-aligned debugging voxel grid")
    voxel_grid = volume.extract_voxel_grid()
    # o3d.visualization.draw_geometries([voxel_grid])

    # print("Extract point cloud")
    # pcd = volume.extract_point_cloud()
    # o3d.visualization.draw_geometries([pcd])

    for i in range(len(camera_poses)):
        print("Deintegrate {:d}-th image into the volume.".format(i))
        color = o3d.io.read_image(rgbd_data.color_paths[i])
        depth = o3d.io.read_image(rgbd_data.depth_paths[i])
        rgbd = o3d.geometry.RGBDImage.create_from_color_and_depth(
            color, depth, depth_trunc=4.0, convert_rgb_to_intensity=False)
        volume.integrate(
            rgbd,
            camera_intrinsics,
            np.linalg.inv(camera_poses[i].pose),
            True
        )
        mesh = volume.extract_triangle_mesh()
        mesh.compute_vertex_normals()
        o3d.visualization.draw_geometries([mesh])