Experimental results show that the proposed algorithm performs well, even for captions with unknown font, size, and position.

• 实验结果表明，即使在复杂背景下，对于字体、大小和位置都不确定的字幕，该算法仍具有良好的分割效果。