Exemplary methods of extracting closed caption (CC) image from a frame of an audio video (AV) stream are described. For all pixels of the frame, setting a color value of the pixels to a determined pixel value when the color value of the pixel is different from a background color value associated with CC image. A set edges is analyzed to identify one or more polygons. A polygon that contains text is determined from the one or more polygons. The frame is cropped along the polygon to obtain a CC image. Upon determination that the CC image is identical to another closed caption image a frame count associated with the other closed caption image is increased by 1; and upon determination that the CC image is not identical to the other CC image the closed caption image is stored along with a position and a time value as metadata information.