基于笔画宽度变换的自然场景文本检测方法

摘要

According to the characteristics that natural scene text tends to maintain a fixed stroke width, this paper puts forward a natural scene text detection method based on the transformation of pixel stroke width. This method calculates stroke width of each pixel in the image by using stroke width transformation, and then merges the adjacent pixels with stroke width approximately equal to form text candidate region. According to the rule it filters part of non text regions so as to realize the text detection. The experimental results show that the proposed text detection method based on stroke width transformation avoids the classification of text characteristics, so that it can improve the precision rate and recall rate of text detection, and shorten the detection time.%　　根据自然场景中文字笔画倾向于固定宽度的特点，提出一种基于像素笔画宽度变换的自然场景文本检测方法。该方法利用笔画宽度变换方法计算图像中每一个像素的笔画宽度，将笔画宽度大致相等的相邻像素合并形成文本候选区域，按照规则滤除一部分非文本区域从而实现文本的检测。实验结果表明，所提出的基于笔画宽度变换的文本检测方法避开了文本特征分类问题，能够提高文本检测的准确率和召回率，同时缩短文本检测的时间。