RVC Camera – Terminologies

Depth Map: An image or image channel that contains information about the distance from the surfaces of scene objects to the viewpoint, used to simulate or reconstruct 3D shapes. Each pixel value in the image represents the distance from a point in the scene to the camera.

Confidence Map: A confidence map is essentially a probability density function over a new image. Based on the color histogram of an object in the original image, a probability is assigned to each pixel in the new image, representing the likelihood of that pixel color appearing in the original object.

Normal Vector: The normal vector of a point cloud can be approximated as a set of perpendicular lines to the surface of the object being photographed. The normal vector computation radius is the radius of a neighborhood sphere around a point in the point cloud. In practice, for each point in the cloud, its k-nearest neighbors are found, and a plane is fitted to these k+1 points. A line through the point and perpendicular to this plane is the normal vector. k is the radius used in the calculation.

Camera Gain: Gain controls the sensitivity of the image sensor to light. The higher the gain, the more sensitive it is. High sensitivity helps under low-light conditions but also amplifies noise, resulting in a lower signal-to-noise ratio.

Gamma Correction: A key nonlinear transformation applied to grayscale values of the input image to correct brightness deviations, often used to enhance dark details. Typically, when gamma > 1, highlights are compressed and shadows are expanded; when gamma < 1, highlights are expanded and shadows are compressed.

HDR (High Dynamic Range Imaging): A set of techniques used to achieve greater exposure dynamic range than standard digital imaging. It works by taking multiple photos of the same scene at different exposure levels and combining them into a single HDR image that can capture a broader range of brightness, from dark shadows to bright highlights.

Clustering: If the distance between two points is less than a certain threshold, they are considered directly reachable. If two points can be connected via a series of directly reachable points, they are indirectly reachable. All reachable points form a cluster. In denoising via clustering, if a cluster contains fewer points than a specified threshold, all its points are considered noise.

Downsampling: Simplifying an image by reducing the number of sampling points in the matrix, often by skipping values. A common method is to take every kth point along each row and column.

White Balance: The foundation for determining color in a camera, based on 18% neutral gray to restore true colors and achieve color balance.

ROI (Region of Interest): A selected area from an image that is to be processed, representing the focus of image analysis.

Camera Extrinsic Parameters: Parameters related to the camera’s position and orientation in the world coordinate system. They include the rotation matrix and translation matrix, which together describe how to convert points from world coordinates to camera coordinates.

Camera Intrinsic Parameters: Parameters related to the camera’s internal characteristics, such as focal length and pixel size.