Module colorization 

Point cloud colorization from registered RGB images.

A “registered” RGB image is one whose camera pose (position and orientation relative to the point cloud coordinate frame) is known. Given the camera intrinsics and extrinsics, every 3-D point can be projected onto the image plane and the corresponding pixel color assigned to it.

Pipeline

1. Transform each 3-D point from world space into camera space using the provided world-to-camera isometry.
2. Project the camera-space point onto the image plane with the pinhole camera model.
3. Discard points that lie behind the camera (negative depth) or whose projected pixel falls outside the image.
4. Sample the pixel color — nearest-neighbour or bilinear — and write it into the output ColoredPoint3f.

Multi-image support

colorize_from_images accepts a slice of (image, intrinsics, pose) triples. Each 3-D point is colored by the first image that can see it. Images should therefore be supplied in order of preference (e.g. most frontal camera first).

Structs

CameraIntrinsics

Pinhole camera intrinsic parameters.

ColorizationConfig

Configuration for colorize_point_cloud.

ColorizationResult

Summary statistics returned by colorize_point_cloud.

RgbImageView

A borrowed, tightly-packed RGB image (row-major, 3 bytes per pixel).

Enums

InterpolationMode

Interpolation strategy when sampling pixel colors.

Functions

colorize_from_images

Colorize a point cloud from multiple registered RGB images.

colorize_point_cloud

Colorize a point cloud from a single registered RGB image.