基于高频小波系数分类器的图像文本信息的非监督检测

Text’s Unsupervised Detection in Image Based on High-Frequency Wavelet Coefficients Classification

图像文本信息的定位与识别在数字图像信息、视频数据库和Web地址的检索应用中十分重要.但文本信息通常镶印在图像的复杂场景中,其检测相当困难.提出了一种能够自动水平校准检测不同大小、字体、颜色和语种的图像文本信息的鲁棒方法.它首先对待测图像进行小波变换,将高频小波系数的分布状况作为文本区与非文本区的统计特征,然后应用K-均值算法分类出图像中的文本区,再经过投影分析以更精确地定位.最后,生成作为OCR引擎输入值的二值文本图像.所提出的检测方法的性能通过试验得到了验证.

Text localization and its recognition in images was important for searching information in digital photo archives, video databases and web sites. However, since it was often printed against a complex background, text was often difficult to be detected. A robust text localization approach was proposed, which could automatically detect horizontally aligned text with different sizes, fonts, colors and languages. First, a wavelet transform was applied to the image and the distribution of high-frequency wavelet coefficients would discrete characterize text and non-text areas. Then, the K-means algorithm was used to classify text areas in the image. The detected text areas under went a projection analysis before refinement their localization. Finally, a binary segmented text image was generated, to be used as input to an OCR engine. The detection performance of the approach was proved experimental demonstration.