Playfield Detection Using Color Ratio and Local Entropy

Playfield detection is one of important tasks in sports video analysis. Although some encouraging results have been achieved, the robustness of playfield detection still needs to be enhanced. To the best of our knowledge, traditional methods for playfield detection usually using various color features. Detection results are unstable when there are shadows on the playfield. In this paper, an effective method for the playfield detection is proposed by fusing the color ratio and local entropy. In the proposed method, two segmentation masks based on color ratio and local entropy are get first. Then, logical AND operation is performed on the two masks. Finally, a morphologically open operation on the binary image is performed to remove the small regions. The experimental results show that the proposed method effectively detects the playfield even with strong shadow.