基于重线和笔画连通性的视频文字二值化方法

文字二值化是实现提取视频文字信息的重要步骤。在复杂背景下,现有二值化方法不仅难以有效抑制背景干扰,而且对文字定位精度要求过高。提出一种复杂背景下稳健的视频文字二值化方法,首先,引入了穿越线的概念,通过找到一条穿越文字行的穿越线,在笔画中植入关键点。进而,充分利用文字笔画的连通性,利用区域生长的方法,完成二值化过程。实验表明,该方法能够在复杂背景下得到较理想的文字二值化图像,不受文字颜色、字体、大小等因素的影响。

An Approach to Video-text Extraction based on Text Traversing line and Stroke Connectivity

Binarization is an important step in video-text retrieval.In complex background,conventional binarization methods could not eliminate interference from background,and further more,imposes an over-strict accuracy requirement on text location.A robust video-text extraction method is proposed in this paper.First,The concept of traversing line is introduced,and the key-points are seeded in strokes by finding a text traversing line.Then,based on the connectivity of strokes,region growth is employed to create a clear binary text image.Experiment indicates that this scheme is robust to the influence of color,font,size,and other factors in complex image.