The Research On The Method Of High-Definition Video Text Extraction And Recognition

Video text extraction and recognition has broad application prospects,such as automobile autopilot,machine pattern recognition and video retrieval,which play an important role.So far,video retrieval are still stuck in the artificial mark when the research of video content has attracted the researchers' attention.To understand the video content,the most available method on the study is to recognition video text.The video text can be divided into the title text and scene text.In this paper,the main text processing steps which text detection and recognition is proposed.The characteristics of the text and the classifier can be able to handle complex text.First of all,based on the current video text research is still in the horizontal direction of text study and video text with high contrast,the paper will process any direction of text;Secondly,in view of the present video text is limited by the diversity of the text characteristic,the low resolution video,problems such as complex background pixels,the paper first uses the gradient vector flow to obtain the boundary of the text and remove most of the background pixel,then uses two classifiers to remove the other background pixel,which can get the position of the text and the text box;After get the text box,cutting processing can get character,which send into the OCR,eventually get the text.Experimental results show that the proposed approach to process the text which in any direction,and get the text.