视频中的文字提取技术.pdfVIP

ABSTRACT To some extent, the high-level semantic information reflects the content of the video, and the text embed in videos contains a wealth of high-level semantic information. If text can be detected, segmented and recognized automatically, there will be a big improvement in high-level semantic information understanding, indexing and retrieval. Text extraction system is normally composed by four parts, namely, text event detection, text area localization, text segmentation and character recognition. This paper concentrates on the algorithm research of text area localization and text segmentation. For text localization, two methods are proposed. The first one is based on wavelet transformation, which applies corner response image and combined high-frequency sub band. And the statistical characteristic is extracted as the feature vector for K-means cluster. Then, some heuristics are employed to remove the false positives. This approach utilizes the classification of unsupervised learning method, therefore, it is able to avoid sample training, which is time-cosuming. The second algorithm is based on Gabor transformation, and it aims at Chinese text detection and localization. The strokes of most Chinese characters are oriented in four directions, so Gabor with various scales and directions can describe it well. The cluster result for different scales of Gabor transformation is combined to obtain the location of text. Experimental results show that the proposed method is robust even in low-contrast condition. For text segmentation, in order to obtain a better recognition result from OCR software, text segmentation part differentiates the text pixels from the background pixels. A color space based segmentation approach i