java - Humble Video 拍摄给定时间的快照

Question

你好，我用的是https://github.com/artclarke/humble-video从视频中获取缩略图。
到目前为止，我已经成功地通过以下方法从视频中拍摄了快照。

private static Path generateThumbnail(final Path videoFile)
     throws InterruptedException, IOException {

 final Demuxer demuxer = Demuxer.make();
 demuxer.open(videoFile.toString(), null, false, true, null, null);

 int streamIndex = -1;
 Decoder videoDecoder = null;
 String rotate = null;
 final int numStreams = demuxer.getNumStreams();
 for (int i = 0; i < numStreams; ++i) {
     final DemuxerStream stream = demuxer.getStream(i);
     final KeyValueBag metaData = stream.getMetaData();
     final Decoder decoder = stream.getDecoder();
     if (decoder != null
             && decoder.getCodecType() == MediaDescriptor.Type.MEDIA_VIDEO) {
         videoDecoder = decoder;
         streamIndex = i;
         rotate = metaData.getValue("rotate", KeyValueBag.Flags.KVB_NONE);
         break;
     }
 }

 if (videoDecoder == null) {
     throw new IOException("Not a valid video file");
 }
 videoDecoder.open(null, null);

 final MediaPicture picture = MediaPicture.make(videoDecoder.getWidth(),
         videoDecoder.getHeight(), videoDecoder.getPixelFormat());

 final MediaPictureConverter converter = MediaPictureConverterFactory
         .createConverter(MediaPictureConverterFactory.HUMBLE_BGR_24, picture);

 final MediaPacket packet = MediaPacket.make();
 BufferedImage image = null;
 MUX : while (demuxer.read(packet) >= 0) {
     if (packet.getStreamIndex() != streamIndex) {
         continue;
     }
     int offset = 0;
     int bytesRead = 0;
     videoDecoder.decodeVideo(picture, packet, offset);
     do {
         bytesRead += videoDecoder.decode(picture, packet, offset);
         if (picture.isComplete()) {
             image = converter.toImage(null, picture);
             break MUX;
         }
         offset += bytesRead;

     } while (offset < packet.getSize());
 }
 if (image == null) {
     throw new IOException("Unable to find a complete video frame");
 }
 if (rotate != null) {
     final AffineTransform transform = new AffineTransform();
     transform.translate(0.5 * image.getHeight(), 0.5 * image.getWidth());
     transform.rotate(Math.toRadians(Double.parseDouble(rotate)));
     transform.translate(-0.5 * image.getWidth(), -0.5 * image.getHeight());
     final AffineTransformOp op = new AffineTransformOp(transform,
             AffineTransformOp.TYPE_BILINEAR);
     image = op.filter(image, null);
 }

 final Path target = videoFile.getParent()
         .resolve(videoFile.getFileName() + ".thumb.jpg");

 final double mul;
 if (image.getWidth() > image.getHeight()) {
     mul = 216 / (double) image.getWidth();
 } else {
     mul = 216 / (double) image.getHeight();
 }

 final int newW = (int) (image.getWidth() * mul);
 final int newH = (int) (image.getHeight() * mul);
 final Image thumbnailImage = image.getScaledInstance(newW, newH,
         Image.SCALE_SMOOTH);
 image = new BufferedImage(newW, newH, BufferedImage.TYPE_INT_BGR);

 final Graphics2D g2d = image.createGraphics();
 g2d.drawImage(thumbnailImage, 0, 0, null);
 g2d.dispose();

 ImageIO.write(image, "jpeg", target.toFile());
 return target.toAbsolutePath(); }

现在，我想做的是在视频开始后 2 秒后拍摄快照，这可能吗？我
已尝试使用“Demuxer”-s seek 方法，但没有运气。

Answer 1

我已经用下面的代码成功地做到了
图书馆的方法

public int seek(int stream_index, long min_ts, long ts, long max_ts, int flags);

参数是

流索引 用作时基引用的流的索引

min_ts 可接受的最小时间戳

ts 目标时间戳

max_ts 可接受的最大时间戳

我的实现是

final int success = demuxer.seek(streamIndex, 0, 700, 99999999,VideoJNI.Demuxer_SEEK_FRAME_get());