H.264

H.264 , or MPEG-4 AVC , is a standard of video coding developed jointly by UIT-T Q.6/SG16 Video Coding Experts Group (VCEG) as well as the ISO/CEI Moving Picture Experts Group (MPEG) and is the product of an effort of partnership known under the name Video Joint TEAM (JVT). Standard UIT-T H.264 and standard ISO/CEI MPEG-4 Part 10 (ISO/CEI 14496-10) are technically identical, and technology employed is also known under the name AVC , for Advanced Video Coding. The first version of the standard was approved in May 2003 and the most recent date of March 2005.

The JVT currently works on the concept of scalability by working out an extension to the standard H.264, it acts of the specifications Scalable Video Coding (SVC).

History

The name H.264 comes from the family of video standards H.26x defined by UIT-T. Within the framework of MPEG, name AVC was by analogy selected with the audio codec AAC MPEG-2 leaves 7 which had been named thus to differentiate it from audio codec MPEG-2 leaves 3. The standard is usually called H.264/AVC (or AVC/H.264 or H.264/MPEG-4 AVC or MPEG-4/H.264 AVC) to underline the common heritage. Name H.26L, pointing out its bond with the UIT-T definitely less common but is always used. From time to time, it is also called “codec JVT”, in reference to the organization JVT which developed it. There exists a precedent in the development of a standard of video coding common between MPEG and UIT-T with MPEG-2 and H.262 which are identical. However, this codec was developed in the framework of MPEG, the UIT-T being satisfied to adopt it then and to publish it in its center.

Objectives and Applications

At the origin, the UIT-T launched project H.26L in 1998 with an aim of creating a new architecture of codec having for goal a profit in effectiveness of coding of a report/ratio at least equal to 2 compared to the existing standards (MPEG-2, H.263 and MPEG-4 Part 2). Another goal was to create a simple interface to be able to adapt the codec to the various protocols of transport (circuit and packet switching). The codec was developed by making sure that it would be transposable on platforms at a reasonable cost, i.e. by holding account of the progress made by industry of the Semi-conducteur S as regards design and of the processes.

In 2001, project H.26L had achieved its goals in compression ratio as subjective tests carried out showed it by… MPEG. It is at this time that the ITU-T and MPEG decided by mutual agreement to create the Video Joint TEAM (JVT) with an aim of standardizing the codec together and to adapt it to the various needs for industry (vidéophonie, streaming, television, mobile). Indeed, the applications traditionally aimed by the ITU-T relate to the low flows (vidéophonie, mobile), applications for which H.26L was optimized, whereas the members of MPEG wished to adapt it to other formats (television, HD). Algorithmic tools as the support of interlaced were added and a reduction of complexity was accomplished.

Codec H.264/AVC is thus adapted to a very large variety of networks and systems (for example, for the diffusion of the Télévision, storage HD-DVD and Blu-Ray, the streaming RTP/IP, and of the systems of Téléphonie suitable for ITU-T).

Following the first version of the standard, the JVT developed some extensions, known under the name Fidelity Range Extensions (FRExt). The purpose of these extensions are to deal with precision of increased quantification (addition of codings 10-bit and 12-bit) and a better definition of the chrominance (addition of the structures of quantification YUV 4:2: 2 and YUV 4:4: 4) and aim at professional applications (Studio). Several other functionalities were also adopted to improve subjective quality in high-definition (addition of a transform 8×8 in addition to the existing transform 4×4, addition of matrices of quantification) or for specific needs (coding without loss, support of other spaces of colors). The work of design on Fidelity Range Extensions was finalized in July 2004, and was solidified in September 2004.

Since the end of the development of the original version of the standard in May 2003, the JVT made publish 4 versions approved by the UIT-T and MPEG, corresponding to the addition of FRExt and corrections.

Detailed characteristics

H.264/AVC (MPEG-4 Part 10) includes/understands many new techniques which enable him to compress the vidéos much more effectively than the preceding standards (H.261, MPEG-1, MPEG-2, MPEG-4 Part 2/ASP) and provides more flexibility to the applications in a great number of environments network. In these principal functionalities are inclus :

a compensation of movement which can be carried out compared to several already coded images of reference. The choice of the image of reference intervenes on the level macroblock and under-macroblock. This makes it possible to use in certain cases up to 32 images of reference (contrary to the preceding standards, which were limited to one or in the case of images B conventional, with two) and up to 4 different references for same a macroblock. This particular functionality allows usually modest improvements the level of the flow and quality in the majority of the scenes. But in certain types of scenes, such as for example the scenes containing of the fast and repetitive flashes or the scenes frequently reappearing, it allows a really significant reduction of the flow.
a compensation of movement which can use 7 different sizes of blocks (16×16, 16×8, 8×16, 8×8, 8×4, 4×8 4×4) allows a very precise segmentation of zones moving.
a precision with the quarter of pixel for the compensation of movement, allowing a very precise description of the displacement of the zones moving. For the chrominance, the precision of the compensation of movement is done even with the eighth of pixel.
a compensation of movement balanced ( Weighted Prediction ) by weights and shifts allowing a coder to build predictions adapting to the change of brightness and chrominance of the current scene. This brings in particular a comprising profit for the scenes of the transitions due to flashes or dissolves between scenes carried out to the assembly.
a filtering anti-blocks ( deblocking filte ), carried out in the loop of coding and operated on the blocks 4×4, allowing to reduce the Artefact S characteristics of coding with transformation into block.
a whole transform carried out on blocks of size 4×4 pixels (near to the traditional DCT). For the new profiles resulting from the FRExt extensions, an additional transform of size 8×8 was added.
a transformed of Hadamard carried out on the average coefficients of the primary space transform (for chroma and possibly the Brightness in certain cases) to obtain even more compression where the image is softened.
a space prediction on the edge of the close blocks for a coding “ intra ” (rather than the only prediction on the continuous coefficients presents in MPEG-2 Part 2 and the prediction on the coefficients of the transform of H.263+ and MPEG-4 Part 2).
a arithmetic Coding (CABAC: Context-adaptive binary arithmetic coding ), which is a sophisticated technique of entropic coding which produces excellent results in term of compression but has a great complexity (nonavailable in the profiles baseline and extended ).
an adaptive coding of type Huffman with variable length (CAVLC: Context-adaptive Huffman variable-length coding ), which is an alternative less complex than CABAC for the coding of the tables of conversion rates. Although less complex than CABAC, CAVLC is more elaborate and more effective than the methods usually used until now to code the coefficients.
a technique simple and highly structured coding with variable length (VLC : Variable length coding ) for many elements of syntax uncoded by CABAC or CAVLC, considered as code exponential Golomb (Exp-Golomb) Exponential-Golomb (Exp-Golomb) code.
a layer of abstraction network (NAL: Network abstraction to bush-hammer ) is defined to allow the use of the same video syntax in many environments network, this includes possibilities such as parameters of sequence (SPS: Sequence parameter set ) and of image (PS: picture parameter set ) which offers more robustness and of flexibility that former designs.
the sections of commutation (called SP and IF) make it possible a coder to direct a decoder so that this last can form part of an entering video stream, this allows video streaming variable flow and an operation in “ trick mode ” (faked mode). When a decoder jumps in the middle of a video stream by using this technique, it can be synchronized with the images present at this place in spite of the use of other images (or not of images) like references preliminary to displacement.
the flexible scheduling of the macroblocs ({in}} FMO: Flexible macroblock ordering, alias slice groups ) and the arbitrary scheduling of sections (ASO: Arbitrary slice ordering ) is techniques of reorganization of the scheduling of the fundamental areas of the image (macroblocs). Typically used to improve resistance to the errors and the losses, these techniques can also be used at other ends.
the partitioning of data (DP: Data partitioning ) gives the possibility of separating the elements from syntax of importance more or less high in various packages from data. This makes it possible to apply an unequal level of protection (UEP: Unequal error protection ) with the errors according to the importance of the data and to improve the reliability of flow thus.
redundant sections (RS: Redundant slices ) makes it possible to improve resistance to the errors and the losses while making it possible to the coder to transmit an additional version of whole or part of the image in a less quality which could be used if principal flow is corrupted or lost.
a simple automated process of prevention against the accidental creation of false codes of starting. They are special binary sequences which are placed within the data, allowing a chance access to the data flow as well as a resynchronisation in the event of temporary loss of flow.
Of information additional of improvement (SEI: Supplemental enhancement information ) and of information of qualitative state of video (VUI: Video usability information ) is extra informations which can be inserted in flow to improve its use for a great number of applications.
Of the auxiliary images can be used for uses such as the mixing by Chanel alpha.
the classification of the images allows the creation of under-sequences (allowing a temporal scalability by optional inclusion of additional images between other images) as well as the detection and the dissimulation of the loss of whole images (which can occur in the event of loss of packages network or errors of transmission).
counting about the images makes it possible to preserve the order of the images and the sound in separately decoded images of information of schedule (what makes it possible a system to transport, control and/or change the information of schedule without affecting the contents of the images).

These techniques, like several others, help H.264 to exceed the preceding standards significantly, in a large variety of circumstances and a large variety of environments of application. H.264 can function often definitely better than it video MPEG-2 by obtaining same quality with a decreased bitrate of half, even more.

Like many other video standards of group ISO/CEI MPEG, the H.264/AVC has a software application of reference, which can be downloaded free (see the section external Liens below).

The main objective of this application is to give examples of the various possibilities of the H.264/AVC, rather than to provide a product really usable and powerful.

A material application of reference is also in the course of standardization by group MPEG.

Profiles

The standard includes the 6 following whole of characteristics, which are called profiles , each one targeting a precise class of applications:

Baseline Profiles (BP) : Mainly for the low-cost applications which use few resources, this profile is very much used in the portable applications and of videoconference.
Main Profiles (MP) : In the beginning, envisaged for the general public applications of diffusion and storage, this profile lost importance when the profile High was added with the same objective.
Extended Profiles (XP) : Envisaged for the diffusion in flow ( streaming ) of the vidéos, this profile has capacities of robustness to the loss of data and change of flow.
High Profiles (HiP) : Principal profile for the diffusion and storage on disc, in particular for high definition television (this profile was adopted for the discs HD-DVD and Blu-Ray like for French digital television high-definition).
High 10 Profiles (Hi10P) : This profile goes beyond the general public applications and is based on the High profile - adding to 10 bits of precision by pixel.
High 4:2: 2 (Hi422P) Profiles: The principal profile for the professional applications, it is based on the profile High 10 - adding the support for the quantification 4:2: 2 to 10 bits per pixel.
High 4:4: 4 (Hi444P) ' Profiles: This profile is based on the profile High 4:2: 2 - adding the support for the quantification 4:4: 4, to 12 bits per pixel and in more the support for a mode without effective loss. Note: the profile High 4:4: 4 is in the course of removal of the standard in favor of a new profile 4:4: 4 under development.

Levels

The levels ( levels ) are limitations on a certain number of parameters which make it possible the decoders to limit the resources memories and calculative necessary to decode a video.

Note : a macrobloc is a zone of 16×16 pixels.

Patents

As for the formats MPEG-2 Shares 1 and 2 like MPEG-4 Share 2, the retailers of products and services using standard H.264/AVC must pay rights for the use of a patented technology. The principal recipient of these rights concerning this standard is a private organization: MPEG it, LLC (which is absolutely not affiliated with the " MPEG standardization organization" , but which also manages patents for systems using MPEG-2 Part 1, of vidéos MPEG-2 Part 2 and MPEG-4 Part 2 as well as other technologies).

Applications

The two principal candidates include the " H.264/AVC High Profile" as an obligatory characteristic of the readers with in particular:

the format HD-DVD of the DVD Forum
the format Blu-ray Disc of Blu-Ray Disc Association (BDA)
the format of video camera Avchd

In Europe, the organization of standardization DIGITAL Video Broadcast (DVB) approved the H.264/AVC for the diffusion of television in Europe at the end of 2004.

The French Prime Minister announced that the H.264/AVC was obligatory in the receivers HDTV and for the paying chains of the Digital terrestrial television (TNT) in France, at the end of 2004.

The organization of standardization Advanced Television Systems Committee (ATSC) with the the United States is considering the use of standard H.264/AVC for the television diffused in the United States.

The Multi-media service DIGITAL Broadcast (DMB) - equivalent with the European TNT - envisaged to be diffused in République of Korea will use format H.264/AVC.

The operators of mobile terrestrial diffusion to the Japan will use the codec H.264/AVC, of which:

NHK
Tokyo Broadcasting System (TBS)
Japanese Television (NTV)
TV Asahi
Fuji TV
TV Tokyo

satellite Direct broadcast TV services will use this new standard, of which:

News Corp. /DirecTV (in the United States)
Echostar/Dish Network/Voom TV (in the United States)
Euro1080 (in Europe)
First (in Germany)
BSkyB (in Great Britain and Ireland)

The 3rd Generation Partnership Project (3GPP) approved the introduction of H.264/AVC like an optional service into version 6 of the functional specifications for the multi-media mobile.

The Motion Imagery Standards Board (MISB) of the organization United States Department off Defense (DoD) adopted H.264/AVC like video codec preferred for all the applications.

The organization Internet Engineering Task Force (IETF) provided a format of setting out of package of contents (RFC 3984) for the transport of video H.264/AVC by using its Real-time Transport Protocol (RTP).

The organization Internet Streaming Media Alliance (ISMA) adopted H.264/AVC for the specifications of ISMA 2.0.

The organization Moving Picture Experts Group (MPEG) integrated successfully the support of H.264/AVC in its standards (for example the systems MPEG-2 and MPEG-4) like in the specifications of the format of file ISO media.

The International union of telecommunications - Sector Standardization (ITU-T) adopted H.264/AVC in the specifications of systems of multi-media telephony H.32x. Based on the standards of the ITU-T, H.264/AVC is already largely used for the videoconference, in particular by two important companies of the market (Polycom and Tandberg). All the new products of videoconference include the support of H.264/AVC henceforth.

H.264 will be probably used in the services of Vidéo to the request on Internet in order to provide Films and television programs on computer. It is also probable that the same type of contents will be proposed via file-swapping on network, in a legal way or not.

Products and implementations

Transposition software

X264 is a H.264 coder distributed under the terms of the license LPG used in the software of transcribing VideoLAN and MEncoder. At the beginning of 2006, x264 is the only transposition complete and free profile Main and High de H.264, interlaced excluded. x264 gained a comparison of video coders organized by Doom9.org in December 2005.

Libavcodec contains a H.264 decoder under license LGPL. It supports the profiles Main and High de H.264, interlaced excluded. It is used in many applications like the readers VLC media player and MPlayer, and the decoders Ffdshow and FFmpeg.

parcelling Nero DIGITAL, Co-developed by Nero AG and Ateme, comprises a decoder and a H.264 coder, and supports other technologies MPEG-4. In September 2005, it supports profiles it Main, interlaced excluded, and should also support soon the profile High and interlaced.

Apple integrated H.264 in Mac OS X version 10.4 (Tiger), like QuickTime version 7, which is provided with Tiger. In April 2005, Apple updated its version of DVD Studio Pro to allow the creation of contents HD. DVD Studio Pro makes it possible to engrave contents HD-DVD at the same time with the standards of media DVD and HD-DVD (even if no engraver is available). To visualize these HD-DVD, one needs the regraver on standard DVD. Apple requires PowerPC G5, Apple DVD Player v4.6, and Mac OS X v10.4 or superior.

the application of videoconference IChat of Apple functions with this codec since version 3 ()

Sorenson offers an application of H.264. The codec Sorenson AVC Pro codec is available in the software Sorenson Squeeze 4.1 for MPEG-4.

Material applications

Several companies produce chips able to decode vidéos H.264/AVC. The chips able to decode in real-time of the vidéos high-definition are inter alia the following ones:

Broadcom BCM7411
Conexant CX2418X
Sigma Designs SMP8634, SMP8635, EM8622L, and EM8624L
STMicroelectronics STB7100, STB7109, NOMADIK (series STn 8800/8810/8815)
WISchip (Micronas the USA, Inc.) DeCypher 8100
Neotion NP4 - NEOTION Processor 4.

This type of chips allows a broad deployment of material low cost able to play of vidéos H.264/AVC to definitions of the standard and high television.

Many materials are right now available in June 2006, that goes from the products of great not very expensive consumption to coders real-time containing FPGA for the diffusion:

Modulus Video delivers coders H.264 real-time of standard TV quality to the diffusers (telephone operators included/understood) and announced its coders real-time high-definition (ME6000) for the semione. The technology of coding Modulus Video HD was shown with the NAB of April 2004 when it obtained the price “Pick Hit”. The Modulus encoding employs the technology of LSI Logic.

Harmonic proposes to a varied range professional video coders supporting H.264 of resolution SD and HD (DiviCom MV 3500 & Electra 5000).

Tandberg television announced a coder real-time high-definition (the EN5990). DirecTV chose this product for its deployment DBS.

At ATI, the series Radeon X1000 of graphics processors comprises a material acceleration of H.264 decoding with the pilots Catalyst 5.13. H.264 decoding is part of multi-media technology " AVIVO3" of ATI [http://www.ati.com/technology/h264.html].

the pilots NVIDIA support material H.264 decoding for certain graphics processors, the list is available on the page PureVideo de NVidia.

the console PlayStation Portable of Sony contains a material decoder of video files to the format H.264

In October 2005, Apple revealed the fifth version of its famous Baladeur IPod able to read vidéos with the H.264 format on its screen or a television. It supports the Baseline profile until the resolution of 640×480, with 30 images a second and with the flow of 768 kbits/s.

WorldGate sells the Ojo cellphones, which use H.264 profiles of it Baseline with the resolution QCIF (176×144) with flows of 80 with 500 kbits/s with 30 images a second.

at the end of 2006, left Archos 604 which manages this format.

External bonds

AVC/H.264, a system of advanced video coding for the HD and the SD, article of the technical review of UER
H.264/MPEG-4 Leaves 10 Tutoriels (Richardson)
H.264/AVC Textbook (in Japanese: Okubo, Kadono, Kikuchi, and Suzuki)
JVT Experts Group document files
MPEG off Terms H.264/MPEG-4 AVC Patent License
MPEG Industry Forum
Page of official publication of the ITU-T
Page of official publication of the ISO

Random links: