H.264 Overview

xiaoxiao2021-03-06  20

1. Introducing the arrival of digital TV and DVD video, has caused a revolution with the home entertainment industry. The standardization of video compression techniques makes these and more applications possible. ITU-t h for video compression 263 Standard is now widely used in video conferencing systems. At the same time, the next standard, MPEG4 in the MPEG series standard is a new generation of Internet-based video applications.

The standard of MPEG4 (visual) and H.263 is based on video compression ("Video Coding") technology. In 1995, it is responsible for the development of these standards - the Motion Picture Experts Group, MPEG and video coding The Video Coding Experts Group, VCEG is in the final stage of new standard development, which has a considerable improvement in this new standard. MPEG4 and H.263 provide better video image compression and support high quality, low-bit rate streaming media video, the new standard, "Advanced Video Coding", has last for at least seven years.

After the H.263 standard for video telephones in 1995, the ITU-T video coding expert group (VCEG) has started further development in two fields, one of which is "short-term" work. Adding additional characteristics to H.263 (the second edition of the standard) and the other for "long-term" work, a new standard for low bit rate visual communications. The long-term work has produced "H. The 26L "standard draft, the draft provides much more video compression efficiency than the previous ITU-T standard. In 2001, the ISO Motion Picture Experts Group, MPEG recognizes the potential of H.26L Benefits. And the Joint Video Team (JVT) has formed a joint video group, which includes experts from MPEG and VCEG. The main task of JVT is to develop the draft of the H.26L "model" into a set of complete sets. International standards. Facts? The results will become two identical standards: ISO MPEG4 Part 10 MPEG4 and ITU-T h.264. New standard "official" named Advanced Video Coding (AVC); however, the standard Work name H.26L and its ITU document number, H.264 [1] is widely known.

2. H.264 CODEC

The same is the same as early standards such as MPEG1, MPEG2 and MPEG4, and the draft The H.264 has not explicitly defined a CODEC (Encoder / Decoder Pair). But define the semantics of the video bitstream and the bit Flow decoding method. However, in practice, the standard encoder and decoder may include the functional elements displayed in Figure 2-1 and Figure 2-2. The function modules displayed in the figure may be subject to standard The desired, and a variety of different forms of CODEC structures should have room for cyclone. The basic functional modules (prediction, transformation, quantization) in the previous standard (MPEG1, MPEG2, MPEG4, H.261, H.263) , Final encoding) is roughly the same, and important changes in H.264 are reflected in the details of each functional module.

The encoder (Figure Figure 2-1) contains two data stream paths, a "forward" road (from left to right, blue) and a "rebuild" path (from right to left, MAGENTA). The data stream path of the decoder shown left (Figure figure 2-2) shows the similarities between the encoder and the decoder.

(Figure)

2.1 Encoder (forward path Forward Path)

Fn represents the input frame to be encoded. This frame is processed in units of macroblock (corresponding to the original image). Each macroblock is encoded in intra mode or Inter mode. No matter which mode, predicting MacRoblock P is based on one Reconstruction frames are formed. In Intra mode, P is formed by samples in the current frame N, which is the front coded, decoded, reconstructed (Uf'n in the figure; note P is not filtered. Sample formation). In Inter mode, P is formed by dynamic compensation prediction of one or more reference frames. In the figure, the reference frame is displayed as a previous encoded frame F'n-1; however, for each A macroblock prediction may be formed by an encoded and reconstructed one or more frames or later frames (in time order). MAcRoblock DN is a minus block or difference block formed by the current MACROBLOCK minus P prediction P. Performing this transformation (using block transformation) and quantization is to give X a set of quantizable transform coefficients. These coefficients are reordered and encoded by Entropy. The factor of entropy encoding, and other information required to decode the macroblock (Such as Macroblock prediction mode, Quantizer step size, motion vector information describing the MACROBLOCK dynamic compensation mode) Forming a compressed bit stream. This bit stream will be sent to Network Abstract Layer, NAL) Transmission or Storage .

2.2 Encoder (Reconstruction Path)

The decoded MACROBLOCK coefficient X is to decode subsequent macroblock to rebuild the frame. The coefficient X has generated a macroblock difference block DN 'with RE-ScALED (Q-1) and inverse transform (T-1). This difference The block is different from the original difference block DN; the quantization process causes the loss of information, so DN 'is a disabled version of DN. To create a rebuilt macroblock uf'n (original MacRoblock's disabled version), to predict MacRoblock P plus To DN '. In order to reduce the adverse effect caused by disability, the filter is used here. After the reference frame is reconstructed from a series of macroblocks f'n.

2.3 decoder

The decoder receives a string compression bit stream from NAL. To generate a quantified coefficient X, the data element is entropy and reordered. DN 'is given by rescaled and inverter (this with the encoder DN 'The same). Use the header information decoded from the bitstream, the decoder creates a predicted Macroblock P, which is the same as the prediction p formed in the encoder. Add P to DN' to generate uf ' n, Uf'n is then generated by the filter to generate a decoded F'n.

From the discussion from the drawing, the reconstruction path in the encoder ensures that the encoder is used to create predictions and decoders to create predictions P. Otherwise, predictive P and predictive P will be made in the encoder. Different, resulting in increments or "drift" between the encoders.

3. References

1 ITU-T REC. H.264 / ISO / IEC 11496-10, "Advanced Video Coding", Final Committee Draft, Document JVTE022, September 200

转载请注明原文地址:https://www.9cbs.com/read-44044.html

New Post(0)