vb.net 从视频中检测字符并提取到图像

Question

example:

file name: test.mp4
file duration: 46min

program will start scan at 0 min to 46 min (frame by frame)
if there is a subtitle or text on video, capture that frame and extract to image

NOT OCR, JUST CAPTURE AND EXTRACT IMAGE

someone told me i can do it by opencv or emgucv
but i dont know how to do it.. so, can you guys give me tutorial?

original image

image after extract

Answer 1

You can just use a video stream in Emgucv, and capture a frame that you want.

private List<Image<Bgr, Byte>> GetVideoFrames(String Filename)
{
        List<Image<Bgr, Byte>> image_array = new List<Image<Bgr, Byte>>();
        Capture _capture = new Capture(Filename);
        double totalFrames = _capture.GetCaptureProperty(Emgu.CV.CvEnum.CAP_PROP.CV_CAP_PROP_FRAME_COUNT);
        double fps = _capture.GetCaptureProperty(Emgu.CV.CvEnum.CAP_PROP.CV_CAP_PROP_FPS);
        double frameNumber = 0.0;

        bool Reading = true;

        while (Reading)
        {
            _capture.SetCaptureProperty(Emgu.CV.CvEnum.CAP_PROP.CV_CAP_PROP_POS_FRAMES, frameNumber);
            Image<Bgr, Byte> frame = _capture.QueryFrame();
            if (frame != null)
            {
                //image_array.Add(frame.Copy());

                //frame.Save(OutFileLocation + "\\" + Guid.NewGuid() + ".jpg");

                frame.Save(OutFileLocation + "\\" + DateTime.Now.ToString("yyyyMMddHHmmssfff") + ".jpg");
            }
            else
            {
                Reading = false;
            }

            frameNumber += (fps * TimeDiff);
        }

        return image_array;
}

If you want to crop the image to only include the captions, then you could just use an roi.

Rectangle rect = new Rectangle(0, 0, frame.Width, frame.Height / 4);
frame.ROI = rect;