AUTOMATIC EXTRACTION OF CLOSED CAPTION DATA FROM FRAMES OF AN AUDIO VIDEO (AV) STREAM USING IMAGE FILTERING

摘要

Exemplary methods of extracting closed caption images from frames of an audio video (AV) stream are described. A first set of frames of a first AV stream including CC images and a second set of frames not including the CC images are received. Each pixel in the first frame is replaced with a dummy pixel, upon determination that a pixel at a corresponding position in the corresponding frame has a same color value to generate a filtered frame including dummy pixels and non-dummy pixels. First coordinates of the top-left most pixel of the first frame that is not a dummy pixel and second coordinates of the bottom-right most pixel of the first frame that is not a dummy pixel are determined. The filtered frame is cropped along the first and second coordinates to extract the CC image.

主权项

1. A method in a processing system that is communicatively coupled to an audio video (AV) source capable of performing closed caption (CC) rendering, of extracting CC images from frames of an AV stream, the method comprising:
receiving from the AV source a first set of frames of a first AV stream, wherein one or more frames of the first set includes CC images; receiving from the AV source a second set of frames of a second AV stream, wherein the second set of frames corresponds to the first set of frames without the CC images; syncing a first frame from the first set with a corresponding frame from the second set; for each pixel in the first frame, upon determination that a pixel at a corresponding position in the corresponding frame has a same color value, replacing the pixel in the first frame with a dummy pixel to generate a filtered frame including dummy pixels and remaining pixels from the first frame that have different color values than pixels at corresponding positions in the corresponding frames; determining first coordinates of the top-left most pixel of the first frame that is not a dummy pixel; determining second coordinates of the bottom-right most pixel of the first frame that is not a dummy pixel; and cropping the filtered frame along the first and second coordinates to extract the CC image.