Caption location and extraction in digital video based on SVM

Text that appears in a scene or graphically added to video can provide an important supplemental source of index information as well as clues for decoding the video´s structure and for classification, and we call them closed caption. In this work, a novel algorithm is presented for detecting and locating caption in digital video. The first module of the system divides an image into small blocks featured by pixel value that is fed to SVM (support vector machine) to classify whether they are text blocks or not. The other module is to do post-processing on the classified text blocks to identify the rectangle region of them and OCR can be used further and easily. Experiments conducted with a variety of video sources show that our method could detect and locate caption region successfully by SVM with comparatively less samples.