A Color-Based Image Segmentation Approach for Traffic Scene Understanding

Abstract

Traffic signs are designed with attractive colors and shapes for human vision. Considering the significant characters, color spaces are analyzed based on amount of experiments. For the specific application of traffic scene understanding, some appropriate color components are recommended in our paper, which can reduce the difficulty of segmentation and satisfy the applications on complicated traffic scenes, such as weather various, illumination changes, complex backgrounds, and blurry conditions. Then an approach using entropy of the histogram is applied to traffic sign segmentation. And the experiment results show the validity of our method.