About-network-video:
video-solution:
Related-links:
|
| H.264 video coding standard analysis of the core technology
|
|
Image Communication is a long way in recent
years the development of modern communications technology, image compression
progress in the development of communication is an important part.
International standards recommended the advent of H.261, the image coding
nearly 40 summing up the outcome of the study and solve the visual
communications technology in the application of this long-troubled people's
problems, covering the entire audio-visual business on narrowband ISDN the
image coding , A strong impetus to the conference TV, TV phones and other
means of communication in international image and industrialization.
Subsequently, ITU H.261 in the proposed proceed on the basis of a very low
rate of image compression standard, developed a proposal H.263, and the
latest H.264/AVC. This article first H.261 and H.263 recommendations on the
basic principle, then the new H.264/AVC standards in the new technologies
that, and then on to sum up H.26x series of standards.
1 H.261 basic principles of the proposed
Each image compression standard-setting, it is most appropriate for
the target application. Is the first definition H.261 video coding standard.
It used for the first time a motion compensation prediction coding and DCT
transform the method of combining its video encoding signal transmission
rate from 64 kbps to 1.92 Mbps, and are thus p ¡Á 64K video encoder (p value
at between 1 to 31). H.261 mainly used the Internet ISDN video conferencing
systems, positioning in the circuit-switched network system.
H.261 encoder principle as shown in Figure 1.
The proposals are mainly used CIF image resolution format and QCIF
resolution format to solve the communication between the different standards
of compatibility problems. For each frame coded slice, H.261 motion
compensation by the Inter prediction algorithm, the elimination of
television images related to the time domain; forecast error on the DCT-domain
to remove images of space on the relevance of the then self DCT coefficient
to quantify, to take full advantage of the visual characteristics; followed
by entropy coding, in order to achieve statistical match coding; final
output buffer memory used to smooth the digital stream, to maintain a
constant rate of the digital output of purpose. The proposed model image
frame coding, including I, P, B three categories. I frame by frame encoding;
P-frame, a frame encoding, I frame or in front of a P-frame motion
compensation, and then encoded on the error estimates; B frame interpolation
frame for the two-way transmission not coded, and by the I frame and the
frame or P P PP frame and the frame interpolation reconstruction. H.261 does
not support two-way movement forecasts and GOP, every frame coding frame
before it is a coded reference frame for frame.
H.261 standard encoding data structure from top to the bottom of the
definition of four levels, namely frame layer, layer, slice layer and block
layer. H.261 valuation of the campaign for compensation is slice of the
units. Acer is a block of choice or intra-frame encoding, first of all need
to judgement. If it matches slice with highly relevant, can be used frame
encoding the contrary, the use of intra-encoding.
2 H.263 basic principles of the proposed
H.263 standard in H. 261 on the basis of criteria proposed. In its
low-rate conditions, can not increase too much complexity of the case, a
higher image quality. In principle, it requires half the bandwidth can be
achieved with the same H.261 video quality. At present, H. 263 standards
have been all kinds of visual telephone terminals widespread adoption of the
agreement.
H. 263 coding standard basic model with the diagram of the structure similar
to H.261 standard. Similarly, the use of image motion compensation forecast
to reduce the time-domain redundancy; compensation to the movement of the
residual market forecasts for discrete cosine transform (DCT) coding; using
variable-length coding (VCL) to quantify the DCT coefficient, and additional
motion vector Information entropy coding.
H. 263 in the H.261 on the basis of recommendations made certain
improvements. Image size using QCIF format, the introduction of a sub-CIF
format, but also allows the use of CIF format. 8 ¡Á 8 using the DCT-Acer
block reunification use the same quantitative step to quantify, can be a
slice using a motion vector, it can also be a slice of each sub-block the
use of a motion vector, therefore, have Block motion compensation capability
and improve the forecast frame. To the motion vector x and y to support the
semi-pixel accuracy motion estimation search window size is restricted to
[-16, +15.5], a differential motion vector prediction coding transmission.
Encoding a two-dimensional forecast and VLC combination of coding, similar
to MPEG-1 standard, all the images into the frame and BP P frame.
H. 263 recommendations to ensure that in the very low rate under the
conditions of a better image quality, H.261 on the basis of mixed code,
adopted the unrestricted movement vector model, syntax-arithmetic coding
mode, the High forecasting model and PB - Frame mode coding technology. In
the unrestricted movement in the cancellation of the vector model as the
benchmark pixel image must be encoded in the region restrictions. Senior
forecasting models in the use of the overlap block motion compensation, but
also to allow movement across the border motion vector. The PB-frame mode, B
frame by frame before a decoding P and the current P-frame decoding a
two-way forecast for reconstruction, thus enhancing the frame rate but did
not significantly increase the number of bits. The above three ways is to
improve the Inter forecast. Based on the grammatical form of arithmetic
coding is used to further reduce the transmission bit rate. In this way, all
of the code of variable length encoding and decoding operations are compiled
using arithmetic operations to replace. To provide these senior coding mode,
can make application in the compression performance and complexity and the
choice between a balanced.
3 H.264 standard of core technology and its characteristics
H.264/AVC is the ITU-T and ISO / IEC joint development of the
latest encoding standards, the first by the ITU-T VCEG made in 1997, the
goal is to put forward a better performance (as opposed to the then H.263 )
Video coding standard.
With the previous compared to some coding standard, H.264 standard
inherited the H.263 video standard and MPEG1/2/4 the merits of the
agreement, but the structure does not change, but in all the major functions
within the module using a number of advanced The technology, improve the
coding efficiency. Its main problems: coding is no longer based on 8 ¡Á 8
blocks, but in the size of a 4 ¡Á 4 faster, the residual transform coding.
Used by the transformation encoding is no longer DCT transformation, but an
integral transform coding. A more efficient use of coding the context of
adaptive binary arithmetic coding (CABAC), also with the corresponding
quantitative process also differentiated. H.264 standard with a simple
algorithm easy to implement, the operator of high precision and without
overflow, computing speed and memory occupied small, weakening the
advantages of block effect, is a more practical and effective image coding
standard.
Here are H.264/AVC standards in the previous standard on the new
technology. H.264 standard image is still used to predict and transform
coding combination of coding structure, the basic structure of the encoder
shown in Figure 2:
The encoder can process the data flow into the former channel to
channel and reconstruction. Enter the code Fn frame, the original image 16 ¡Á
16 pixel block Acer encoded. Slice into intra-coding and coding frame
coding. In any case, Acer forecast by the P block access to reconstruction
frame. In intra-coding mode, P from the current frame of encoded slice the
decoding, reconstruction was forecast, such as the figure above the uF'n.
Encoded in the frame mode, P by one or more reference frame motion
compensation, as was forecast, such as F'n-1. Acer forecast to block P Fn
with the current slice of the margin as a residual slice Dn, the
transformation, be quantified after a string of transformation parameters X.
X parameters of the need for two aspects of the handling, sorting and
re-First, entropy transform, the whole process no feedback component, known
as the former is to channel the anti-quantify and inverse transform, a slice
D'n, and then Acer block P are the sum of reconstruction slice uF'n, to be
rebuilt after a series of handling the reference frame F'n, for the next
frame motion estimation, known as the reconstruction of access.
3.1 intra prediction model code
In video encoding, the usual method is the whole image is divided
into a number of macro blocks, and then on each slice encoded. Intra coding
used in the Inter or two models. In Intra mode usually carried out directly
on the slice DCT transformation, the transformation coefficients entropy
coding. To do so to a certain extent, the elimination of intra-space
redundant, but the DCT only a slice of the internal correlation between the
pixels, without taking into account the adjacent slice the correlation
between. H.264 introduction of the Intra forecast methods, the use of
adjacent slice of treatment related to the macro block coding forecast, the
forecast error to transform coding, to eliminate redundant space. It is
worth noting that before the standard is in the forecast transform domain,
and H.264 is directly in the space domain to predict.
3.2 Inter prediction coding mode
H.264 in motion estimation of many new mining technology, including
variable block size, multi-frame motion estimation, the accuracy of
sub-pixel motion estimation and the effect of filtering to block.
¢Å filter to block effect
It is used to eliminate the role of images in the decoding of the
block. Block of the reasons is the slice separately quantified, so in the
adjacent block at the junction of Wang, because of different quantitative
step which is very close to the original value of the pixel after the
Reconstruction of a larger difference, a clear Block border. To block the
effect of filtering is 4 ¡Á 4 blocks on the border filtering, to block the
border tends to smooth.
¢Æ variable block size
Block size of the estimated effect of the campaign is influential.
Acer will be separated into different size pieces of the motion compensation
sub-block structure known as the tree motion compensation. Slice of the
division and partition of the sub-slice, including all four types, as shown
in Figure 3. Smaller pieces can make more accurate estimates Movement, a
smaller movement residuals, the lower rate. H.264 recommendations in the
choice of different sizes in the block, we can see that a slice can carry up
to 16 different motion vector. With multi-frame motion estimation, the same
slice of the different pieces can also use a different frame of reference to
the forecast.
Figure 3 Motion Compensation slice segmentation
Top: slice of the division
At the bottom: slice of the sub-division
¢Ç multi-frame motion estimation
Previous video compression standards used in the campaign Danzhen
estimated technology compared to, H.264 use of the multi-frame motion
estimation has a higher efficiency, greater stability of the error. The
so-called multi-frame motion estimation is that the use of one or more
reference frame to estimate the motion vector, can prevent errors in a frame
and the frame behind affected. However, this estimate needs more memory and
higher computing complexity.
¢È sub-pixel accuracy motion estimation
In H.264, the campaign is estimated by the accuracy of H.263-pixel
to pixel, and the pixels as optional. And the accuracy of the semi-pixel
motion estimation, the accuracy of the pixel motion estimation by the use of
interpolation pixels and pixel-location of the point.
H.264 forecast in the frame coding, can continue to use three-step
search algorithm to identify and match the current macro block by block. In
Block, block and block the displacement of the center or block any point is
equivalent to the displacement. Therefore, the block of displacement can be
understood as the center of displacement. In the three-step algorithm, the
search for 7, that is, in the last one to the current sub-block for the
original point, the current sub-block in its distance from top to bottom
about seven within the scope of certain rules of movement, each move to a
location, took out the same Sub-block size of the current sub-block with the
match calculation. Specific divided into the following three steps:
¢Ù to the current sub-block as the center and 4 for stride, the
successful bidder will Figure 4 to 9 as the central location of the
sub-block with the current sub-block match, sought the best match of the
sub-block center.
¢Ú to ¢Ù obtained in the best sub-block as the center, for example, x
= 4, y = 0, 2 for stride, will map the nine position as the center of the
sub-block with the current sub-block match, obtained The best match of the
sub-block center.
¢Û to ¢Ú obtained in the best sub-block as the center, for example, x
= 4, y = 0, 1-stride, will map the nine position as the center of the
sub-block with the current sub-block match, obtained The best match of the
sub-block center, with the current sub-block center of which is offset the
estimated displacement.
3.3 integral DCT -
H.264 standard used in the 4 ¡Á 4 DCT-rounded as a residual slice of
the basic transformation, this transformation has been the target of motion
compensation predict or intra prediction after the data contained residuals
of the 4 ¡Á 4 block. Such transformation is based on the DCT transformation,
but different from the DCT.
As DCT transformation is real, quantifiable, to be rounded to the
coefficient, thus affecting the accuracy of the operator. At the same time,
the traditional DCT there does not match the problem, generate a reference
frame of the migration, a direct impact on the quality of image
reconstruction.
H.264 proposed integral DCT transform the use of all the operations
are integral algorithm, the core of the transformation is mainly in northern
and displacement. Throughout the process of transformation and quantified,
only the implementation of 16 bit integer algorithms and a multiplication
operation. As long as the H.264 recommendations on the basis of the proper
use of the appropriate anti-change, encoders and decoders will not appear
does not match the phenomenon. The pros and cons of transformation matrix,
respectively
One of the factors are basically rounded, 1 / 2 can be used instead
of displacement. As in the transformation of the multiplication can be
replaced by the transfer operation, therefore, reduce the complexity, but
also solved the problem of accuracy.
H.264 slice of the size of 16 ¡Á 16, each of which 4 ¡Á 4 blocks the
size of the above 4 ¡Á 4 transform the DCT, received 16 4 ¡Á 4 of the
transformation matrix. To further improve the compression efficiency, the
proposal also allows each of the 4 ¡Á 4 of the transformation matrix in the
DC component DC, took out a separate component of a new 4 ¡Á 4 matrix, this
matrix Hardamard transformation. Slice the data transmission sequence shown
in Figure 5.
3.4 entropy coding
H.264 also adopted the recommendations of two entropy coding model:
Based on the context of binary arithmetic coding CABAC, and variable length
encoding VLC. VLC code also includes context-based adaptive variable length
encoding CAVLC.
CABAC use of the arithmetic coding method, a symbol can be less than
1 bit to that. According to the assumption that no error test under the
condition of the data collected indicates that the rate of all, CABAC
performance are stronger than CAVLC. But the anti-CAVLC error of stronger
than CABAC, the complexity of computing and also far below the CABAC.
Therefore, H.264 Baseline Profile provisions in the use of CAVLC, and the
use of the Main Profile for CABAC entropy coding.
4 Summary
In the past compared to the video coding standard, H.264
recommendations in its structure, motion estimation and motion compensation,
slice and quantify the change and entropy coding, and other areas has
significantly improved, with higher coding efficiency and Stronger network
adaptability. In the same image quality under, H.264/AVC algorithm than the
previous standards such as H.263 or MPEG-4 saves about 50 percent of the
rate. Profile of different H.264 can be applied to both real-time
communication, but also can be used to delay Yaoqiubugao other applications.
In addition, the proposed increase in the NAL layer, will be responsible for
encoding the output stream to various types of network adapter, which the
network transmission has better support. At the same time, it has strong
anti-BER characteristics, can adapt to packet loss rate is high and a
serious interference in the wireless channel video transmission. Therefore,
H.264 support different network resources under the classification code
transmission, to obtain a smooth image quality, can be adapted to different
networks in the video transmission, network affinity good.
In today's Internet, multimedia services to the needs of the present
rapid growth trend. Because of the wireless network bandwidth resources and
limited transmission capacity constraints, currently on the market end-users
are mostly paid in accordance with the flow approach to the use of wireless
network data services, increased compression efficiency is a wireless video
and multimedia applications the main objective. So H.264 / AVC coding
standard as the Multimedia Messaging Service (MMS), packet-switched
streaming services (PSS) and the conversations of the most competitive
candidate standards. At the same time, H.264/AVC no restrictions on
ownership, is a public open standards. Therefore, the strengthening of the
various manufacturers in the low-cost manufacturing process in the
competition, allowing for the rapid decline in prices, this technology can
allow for more people to services.
¡¡
Recent Post:
My Son Could Have Been Alive If I Had A Wireless Hidden Camera
Who watches the watchers in surveillance society?
Security Cameras Move Into Houma Public Housing
How IP-based video surveillance works -- way beyond analog CCTV
¡¡
IP/Network camera server/Wireless Network
Camera/Knowledge Articles:
|
| |
|
|
