Extraction of text from videos based on layered method

The text data present in video contain certain useful information for automatic annotation, indexing, and structuring of images. However variations of the text due to complex background make the problem of automatic text extraction extremely difficult job. To address this problem a large number of techniques have been developed. Here a technique to extract text from video is proposed. Video can be considered as a sequence of frames. Identify a frame that contains text information. Take that particular frame as image. Then color clustering is used to divide images into multiple layers according to main color class. After that, obtain the candidate text regions from each color layer and determine whether the candidate text regions is real text regions in the corresponding image layer. The interference of the background can be effectively minimized, since the monochrome color exists in each layer, which can significantly increase the accuracy of text regions localization. After wards, an OCR package is used to recognize the text regions. Since the text region has only one color, it helps to greatly increase the recognition rate. Finally, the relationship between different layers is used to verify the recognition results by the text location.