中南民族大学学报（自然科学版）2018，Vol.37Issue(1)：73-79,7.
笔画背景抑制的自然场景文本检测
Text Detection in Natural Scene Imagebased on Stroke-Background Suppression
摘要
Abstract
Aiming at the problem that the excessive image edge information and noise in complex background causes Stroke width detection's inaccuracy, a new Stroke Width Transform(SWT)text detection method based on texture background suppression was proposed. We use Butterworth's high-pass filtering and the texture features to achieve the preserving the information of the text area while suppressing the background. The application of weighted guide filtering image enhancement technique reduced the interference of the noise in the edge detection,and effectively improved extraction of the stroke width and the edge gradient information's accuracy. Consequently, the performance of SWT text detection has been improved. Experimental results based on ICDAR database verify the effectiveness of this proposed method.关键词
文本检测/笔画宽度变换/高通滤波/纹理特征Key words
text detection/stroke width transform/high pass filter/texture feature分类
信息技术与安全科学引用本文复制引用
熊承义,田清越,高志荣,龚忠毅..笔画背景抑制的自然场景文本检测[J].中南民族大学学报（自然科学版）,2018,37(1):73-79,7.基金项目
国家自然科学基金资助项目(61471400) （61471400）
