Parameter optimization of the SAD-IGMCT for stereo vision in RGB and HSV color spaces

Dependable 3D perception modules are essential for safe operation of robotic platforms. Furthermore, robot navigation and localization as well as object recognition tasks also require processing 2D color camera images. This information could be synchronously delivered by stereo vision sensors with the 3D information automatically mapped onto the 2D camera image. However, embedded real-time stereo vision sensors are often restricted to grayscale images due to limited computational resources. Therefore, we present how a previously designed realtime stereo matching algorithm, the SAD-IGMCT, can be optimized for the RGB and HSV color spaces, further reducing the algorithm´s complexity, while still allowing for a high accuracy.