Color image segmentation based on thresholding for advanced driver assistance systems

Many Advanced Driver Assistance Systems (ADASs) rely on an image obtained by a camera that is mounted on a vehicle. To get useful information in real-time, the acquired image is processed with different computer vision algorithms running on the vehicle's embedded platform. The common preprocessing task is the image segmentation based on color which is often used in lane detection or traffic light/sign recognition algorithms to extract key regions. In this paper, we focus on a color image segmentation based on thresholding. While being very simple, its effectiveness largely depends on the details of the implementation such as the chosen color space or characteristics of the used embedded platform. We provide details regarding PC implementation and ADAS development board implementation as well as the details regarding optimizations that are carried out to achieve smaller execution time on PC and board's Texas Instruments TDA2xx System-On-Chip. The image segmentation mean processing time is reported for three different resolutions and three different color models (RGB, HSV, YUV) for both PC and ADAS development board. The obtained results can help in planning and allocating resources on the vehicle's embedded platform for such computer vision tasks.