A New Shadow Removal Algorithm Based on Susan and CIELAB Color Space
Haiying Zhang, Qirong Zheng, Guiwen Zheng · 2014
For on-road vehicle detection, the self-shadow of vehicle is a great disturbance of accurate detection. Consequently, the detection and removal of vehicle shadow is a primary task for video vehicle detection. For the present shadow removal algorithms which based on different color spaces, miss detection will happen while dealing with the vehicles having the similar color with their shadows. So that, in this paper a new scheme based on CIELAB color space is given and by this way, the limitation to the faint color is overcome by its inherent sensitive to the luminance. At the same time, the shadow removal can be optimized by using Susan operator, which based on the distinguished texture feature between the vehicle and its shadow. The experiments showed that it has robust performance for the vehicles with different colors, especially black ones, which can achieve a higher detection ratio compared with other methods.