GDR(Gradual Decoding Refresh, GradualDecoder Refresh)

GDR(Gradual Decoding Refresh, GradualDecoder Refresh)逐渐刷新，是相对于IDR(Instant Decoder Refresh)的一个概念。为了得到更好的用户体验和更好的网络友好性能，才搞了这么一个Feature.

 

I-frames rely only on their own internal data and they enable decompression or decoding at the start of the sequence. I-frames are compressed using the intra-frame method. I-frames are also known as key frames because their content is independent of any other frames and they can be used as a reference for otherframes. As discussed in the inter-frame compression method, a key frame orinitial frame is used at the beginning of the sequence of pictures to becompressed.

Instant Decoder Refresh (IDR) frames, Gradual DecoderRefresh (GDR) frames as shown in the picture,

Long-TermReference Picture (LTRP) frames are well known I-frames

The main difference between IDR and GDR frames is that a GDR framecan be divided into smaller frames and sent at smaller time intervals whereasan IDR frame is sent in one packet. The purpose of using GDR frames is to avoid a significant surge in the data rate that occurs with IDR frames and to provide a better experience of video quality for the end user. GDR, which allows the images to be refreshed gradually over a small period. Rather than refreshing the whole Image all at once with an IDR frame。

For example, a GDR implementation can send 10 individual sections of a complete frame, and each of these sections is itself an IDR encoded video picture. Because only 1/10 of the frame is gradually changing over a 10-frame window, the user perception of the video quality is generally very good.

LTRPframes, on the other hand, are part of the media resilience provisions that some codecs implement. In evitable network loss and errors of compressed video cause visual errors at the decoder. The errors would spread across subsequent P-frames.  An obvious way to avoid this problem is to have the decoder request an I-frame from the encoder to flush out the errors (error feedback mechanism). However, a better way is to employ different reference frames (older, long-termframes). The feedback mechanism, in conjunction with the known good LTRP frame, helps to repair the lost video data (for example, slices) and to phase out the bad LTRP frame. In codecs that support this implementation, the LTRP frame is the last I-frame that arrived at the codec (using either IDR or GDR) The receiving codec then stores this frame as the LTRP frame. When a new I-frame arrives, it is promoted to LTRP, and so on. If an I-frame is lost during transmission, the receiving codec attempts to use the LTRP frame to recover.

I-frames are compressed using the intra-frame technique, and this has a direct impact in the bandwidth consumption of the video stream. The more frequently I-frames are used, the more bandwidth is required.

 

目前用vlc可以正常解码有GDR feature的Stream,但是好像quicktime则不行。

谁晓得GDR是不是已经成为了H264/5的标准了呢？