Text Detection And Extraction Technique In Video Frame

With the computer and communications technology,broadband network technology,audio and video compression technology as well as the development of computer hardware technology, image and video information of the text for indexing,retrieval and the automatic high-level semantic understanding is very important. This paper mostly studies how to test,extract and recognize the artificial additives text information from video frames. The main job of thesis is as follows:Firstly, according to the characteristics of text and the current text localization and extraction algorithm in video frame which a method base on edge detection text positioning is realized. The main steps include that gray-scale process, edge detection, edge image binarization and mathematical morphology operation in video frame; finally we position the area of text. Experiment shows that the method is easy and can position most of the area of text, but video frame contains an enrich information of the edge in the background, and positioning the text area is not accurate enough.Secondly, we extract and recognize the good text area that the threshold-based segmentation method is used to extract the text. When we deal with the text in a simple or single background video frame, we make use of the Otsu method (OTSU) which is classical threshold segmentation algorithm and the method is simple,robust and effective. When extract text in a complex background video frame, we make use of local adaptive threshold method. Experimental shows a very good segmentation results.Finally, we wipe off noise to the extracted text and maximize remove the background of text, we provide a complete definition of binarization input image for the OCR recognition system. We use the"hold the 7"text recognition software to test text image, the result shows that correct recognition rate of characters can achieve about 80%.