CN105898334B - A kind of DC prediction circuits and its method applied to coding and decoding video - Google Patents
A kind of DC prediction circuits and its method applied to coding and decoding video Download PDFInfo
- Publication number
- CN105898334B CN105898334B CN201610471667.9A CN201610471667A CN105898334B CN 105898334 B CN105898334 B CN 105898334B CN 201610471667 A CN201610471667 A CN 201610471667A CN 105898334 B CN105898334 B CN 105898334B
- Authority
- CN
- China
- Prior art keywords
- sum
- value
- calculating
- block
- calculated
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/70—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals characterised by syntax aspects related to video coding, e.g. related to compression standards
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/102—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
- H04N19/103—Selection of coding mode or of prediction mode
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/46—Embedding additional information in the video signal during the compression process
- H04N19/463—Embedding additional information in the video signal during the compression process by compressing encoding parameters before transmission
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
- H04N19/593—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving spatial prediction techniques
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
A kind of DC prediction circuits and DC Forecasting Methodologies applied to coding and decoding video of the invention;It is characterized in that including adder Multiplexing module, median register module, state self-adaption machine control module;The maximum that adder Multiplexing module obtains resource by being multiplexed adder effectively utilizes;Median register module is computed repeatedly by the way that important median storage is passed to the next cycle as input to reduce;The transmission of state self-adaption machine control module control data and redirecting for calculating cycle;The present invention can reduce circuit work area, reduce the execution cycle of circuit, increase working frequency and improve the accuracy rate of calculating on the basis of DC prediction algorithm functions are realized.
Description
Technical field
The invention belongs to the predictive coding field of video coding and decoding technology, specifically a kind of video that is applied to is compiled
The DC prediction circuits and its method of decoding.
Background technology
With current scientific and technical continuous development, information technology and computer internet are changing in various degree respectively
The daily life of people.Nowadays, people obtain information and are mainly derived from multimedia messages, and multimedia messages are with video
For core.During video storage and transmission are constantly widely used, multimedia technology starts to expand research field gradually
It is charged in the technology of efficient video coding.People for HD video and the requirement that occupies little space it is more and more stronger, video
Encoding and decoding more seem important.DC predictive codings are one of intraframe prediction algorithms, and its popularity uses, and also cause people
To the performance requirement more and more higher of its hardware circuit.
In the prior art, the hardware circuit of DC predictive codings has been greatly improved, and Zhou Wei, Huang Xiaodong et al. exist
2013 PCS (Picture Coding Symposium) deliver " EFFICIENT INTRA PREDICTION VLSI
Circuit described in ARCHITECTURE FOR HEVC STANDARD ", compared in current DC prediction circuits and DC Forecasting Methodologies
Simple and quick circuit;But still suffer from following four problem:
1st, circuit area not enough optimizes, and three-input adder is used for multiple times, and calculating is repeatedly duplicated in calculating process;
2nd, there is the defects of result of calculation is not accurate enough in prediction circuit, and there be the space reduced in the algorithm cycle, and cycle length is drawn
The low processing speed of data, so that all having an impact to the speed of whole coding and decoding video;
3rd, the working frequency 357MHz of circuit, has greatly improved relative to circuit before, but still waits to improve to adapt to
High-frequency coding rate, if the frequency of DC prediction algorithms is bottleneck, it will drag down the real-time coding speed of whole video sequence;
4th, there is the problem of inaccurate in implementation method, fail to consider in DC prediction algorithms completely, for caused by displacement
Decimal place directly abandons problem, and accuracy rate reduces.
The content of the invention
The present invention proposes a kind of DC applied to coding and decoding video to solve above-mentioned the shortcomings of the prior art
Prediction circuit and its method, to reduce circuit work area, reduce circuit computing cycle, raising working frequency and calculating
Accuracy rate, so as to reduce the cost of coding and decoding video and power consumption.
The present invention is that technical scheme is used by reaching above-mentioned purpose:
A kind of the characteristics of DC prediction circuits applied to coding and decoding video of the present invention, is, for any one in 2 × N cycle interior predictions
Pixel value in individual image in 4N × 4N regions, it is a prediction block to remember 4N × 4N regions
Then top a line reference pixel reconstructed value of the prediction block PU is designated as A=[R0,-1,R1,-1,…,R4N-1,-1], the prediction block
The PU row reference pixel reconstructed value of left one is designated as L=[R-1,0,R-1,1,…,R-1,4N-1];It is T to remember the cycle, then TmRepresent
M-th of cycle;Initialize n=1, m=1;
The DC prediction circuits include:Adder Multiplexing module, median register module, state self-adaption machine control mould
Block;
The adder Multiplexing module is in m-th of cycle TmIt is interior to top a line reference pixel reconstructed value A and left
One row reference pixel reconstructed value L carries out n-th calculating, obtains 3 accumulated values SUM_0, SUM_4 and SUM_6 of n-th calculating simultaneously
Pass to the median register module to be stored, and obtain the first sub-block of the prediction block PU when n-th calculatesAnd directly export;
The state self-adaption machine control module reads 3 that the n-th calculates from the median register module
Individual accumulated value SUM_0, SUM_4 and SUM_6 simultaneously pass to the adder Multiplexing module;
The adder Multiplexing module is in the m+1 cycle Tm+1The interior three accumulated value SUM_0 calculated the n-th,
SUM_4 and SUM_6 and top a line reference pixel reconstructed value A carries out n-th calculating, obtains described in when n-th calculates
Prediction block PU the second sub-block PU "n=[P0,0,P1,0,…,P4n-1,0];To the n-th calculate three accumulated value SUM_0,
SUM_4 and SUM_6 and the row reference pixel reconstructed value L of the left one are calculated, and obtain the prediction when n-th calculates
Block PU the 3rd sub-block PU " 'n=[P0,1,P0,2,…,P0,4n-1];
The first sub-block PU ' when being calculated by the n-thn, the second sub-block PU "nAnd the 3rd sub-block PU " 'nForm n-th
Prediction block PU during calculatingn, n+1 is assigned to n, after m+2 is assigned into m, repeats the n-th meter of the adder Multiplexing module
Calculate, untill n=N and m=2 × N, so as to obtain prediction block PU.
The characteristics of DC prediction circuits of the present invention applied to coding and decoding video, lies also in, and the adder is multiplexed mould
Block includes:No.1 adder, No. two adders, No. three adders, No. four adders, No. five adders, No. six adders, No. seven
Adder, No. eight adders, No.1 shift unit, No. two shift units, No. three shift units, No. four shift units, No. five shift units, No. six
Shift unit, No. seven shift units, No. eight shift units, No. nine shift units;
The n-th of the adder Multiplexing module is calculated as:
In m-th of cycle TmInterior, the No.1 adder to No. four adders is rebuild to top a line reference pixel
The value A and row reference pixel reconstructed value L of left one carries out n-th calculating, obtains four accumulated values SUM_0, SUM_ of n-th calculating
1、SUM_2、SUM_3;
Four accumulated values SUM_0, SUM_1, SUM_2, SUM_3 being calculated by No. eight adders the n-th and
Constant " 4 " is calculated, and obtains the accumulated value SUM_7 of n-th calculating;
The No.1 shift unit carries out moving to right 3 bit manipulations to the accumulated value SUM_7 that the n-th calculates, and obtains n-th meter
Calculate average value DCvalue;
No. two shift units carry out moving to left 1 bit manipulation to the average value DCvalue that the n-th calculates, and obtain n-th
The median DCvalue ' of calculating;
The average value that the median DCvalue ' and n-th that No. six adders are calculated the n-th are calculated
DCvalue is calculated, and obtains the accumulated value SUM_5 of n-th calculating;
The accumulated value SUM_5 and constant " 2 " that No. seven adders are calculated the n-th are calculated, and obtain n-th
The accumulated value SUM_6 of calculating;
The median DCvalue ' and constant " 2 " that No. five adders are calculated the n-th are calculated, and obtain
The accumulated value SUM_4 of n computations;
The first sub-block PU that the average value DCvalue calculated using the n-th calculates as the n-thn' in it is all
Element value;
In the m+1 cycle Tm+1Accumulated value SUM_0 and n-th interior, that the No.1 adder is calculated the n-th
The SUM_4 of calculating carries out n-th calculating, and obtained result recycles No. three shift units to carry out moving to right 2 bit manipulations, obtains the
Second sub-block PU " of n computationsnIn first element value Pn-1,n-1;
No. two adders are to the element value R in top a line reference pixel reconstructed value An,-1Calculated with n-th
Accumulated value SUM_6 calculated, obtained result recycles No. four shift units to carry out moving to right 2 bit manipulations, obtains n-th
The the second sub-block PU " calculatednIn second element value Pn,n-1;
No. three adders are to the element value R in top a line reference pixel reconstructed value An+1,-1Calculated with n-th
Accumulated value SUM_6 calculated, obtained result recycles No. five shift units to carry out moving to right 2 bit manipulations, obtains second
Sub-block PU "nIn the 3rd element value Pn+1,n-1;
No. four adders are to the element value R in top a line reference pixel reconstructed value An+2,-1Calculated with n-th
Accumulated value SUM_6 calculated, obtained result recycles No. six shift units to carry out moving to right 2 bit manipulations, obtains n-th
The the second sub-block PU " calculatednIn the 4th element value Pn+2,n-1;
No. five adders are to the element value R in the row reference pixel reconstructed value L of left one-1,nCalculated with n-th
Accumulated value SUM_6 calculated, obtained result recycles No. seven shift units to carry out moving to right 2 bit manipulations, obtains n-th
The 3rd sub-block PU " ' calculatednIn first element value Pn-1,n;
No. six adders are to the element value R in the row reference pixel reconstructed value L of left one-1,n+1Calculated with n-th
Accumulated value SUM_6 calculated, obtained result recycles No. eight shift units to carry out moving to right 2 bit manipulations, obtains n-th
The 3rd sub-block PU " ' calculatednIn second element value Pn-1,n+1;
No. seven adders are to the element value R in the row reference pixel reconstructed value L of left one-1,n+2Calculated with n-th
Accumulated value SUM_6 calculated, obtained result recycles No. nine shift units to carry out moving to right 2 bit manipulations, obtains n-th
The 3rd sub-block PU " ' calculatednIn the 3rd element value Pn-1,n+2。
A kind of the characteristics of DC Forecasting Methodologies applied to coding and decoding video of the present invention be for 2 × N cycles interior prediction any one
Pixel value in image in 4N × 4N regions, it is a prediction block to remember 4N × 4N regions
Then top a line reference pixel reconstructed value of the prediction block PU is designated as A=[R0,-1,R1,-1,…,R4N-1,-1], the prediction block
The PU row reference pixel reconstructed value of left one is designated as L=[R-1,0,R-1,1,…,R-1,4N-1];It is T to remember the cycle, then TmRepresent
M-th of cycle;N ∈ (1,2,4,8,16);1≤n≤N;1≤m≤2×N;
The Forecasting Methodology is to carry out as follows:
Step 1, initialization n=1, m=1;
Step 2, in m-th of cycle TmIt is interior, obtain the accumulated value SUM_0 of n-th calculating using formula (1):
SUM_0=Rn-1,-1+R-1,n-1 (1)
Step 3, in m-th of cycle TmIt is interior, the average value DCvalue of n-th calculating is obtained using formula (2), and with described
The first sub-block that average value DCvalue calculates as the n-thIn all members
Element value;
Step 4, in m-th of cycle TmIt is interior, obtain the accumulated value Sum_4 of n-th calculating using formula (3):
Sum_4=2 × DCvalue+2 (3)
Step 5, in m-th of cycle TmIt is interior, obtain the accumulated value Sum_6 of n-th calculating using formula (4):
Sum_6=3 × DCvalue+2 (4)
Step 6, in the m+1 cycle Tm+1It is interior, obtain the second sub-block PU " of n-th calculating using formula (5)n=[P0,0,
P1,0,…,P4n-1,0] in first element value Pn-1,n-1:
Pn-1,n-1=(SUM_0+SUM_4) ÷ 4 (5)
Step 7, in the m+1 cycle Tm+1It is interior, obtain the second sub-block PU " of n-th calculating using formula (6)nIn second
Individual element value Pn,n-1:
Pn,n-1=(Rn,-1+SUM_6)÷4 (6)
Step 8, in the m+1 cycle Tm+1It is interior, obtain the second sub-block PU " of n-th calculating using formula (7)nIn the 3rd
Individual element value Pn+1,n-1:
Pn+1,n-1=(Rn+1,-1+SUM_6)÷4 (7)
Step 9, in the m+1 cycle Tm+1It is interior, obtain the second sub-block PU " of n-th calculating using formula (8)nIn the 4th
Individual element value Pn+2,n-1:
Pn+2,n-1=(Rn+2,-1+SUM_6)÷4 (8)
Step 10, in the m+1 cycle Tm+1It is interior, obtain the 3rd sub-block PU " ' of n-th calculating using formula (9)nIn
One element value Pn-1,n:
Pn-1,n=(R-1,n+SUM_6)÷4 (9)
Step 11, in the m+1 cycle Tm+1It is interior, obtain the 3rd sub-block PU " ' of n-th calculating using formula (10)nIn
Second element value Pn-1,n+1:
Pn-1,n+1=(R-1,n+1+SUM_6)÷4 (10)
Step 12, in the m+1 cycle Tm+1It is interior, obtain the 3rd sub-block PU " ' of n-th calculating using formula (11)nIn
3rd element value Pn-1,n+2:
Pn-1,n+2=(R-1,n+2+SUM_6)÷4 (11)
So as to obtain the second sub-blocks of prediction block PU PU " when n-th calculatesn=[P0,0,P1,0,…,P4n-1,0] and the
Three sub-block PU " 'n=[P0,1,P0,2,…,P0,4n-1];
Step 13, the first sub-block PU when being calculated by the n-th1', the second sub-block PU "nWith the 3rd sub-block PU " 'nForm
Prediction block PU when n-th calculatesn;
Step 14, n+1 is assigned to n, after m+2 is assigned into m, repeat step 2 performs, until n=N and m=2 × N are
Only, so as to obtaining prediction block PU.
Compared with prior art, advantageous effects of the invention are embodied in:
1st, the existing DC prediction algorithms circuit design of optimization proposed by the present invention, overcomes that circuit footprint is big, work frequency
The problem of rate is not high enough, it is proposed that a kind of hardware implementation mode of DC prediction algorithms applied to coding and decoding video, employ point
Cut processing framework, be multiplexed by adder, state self-adaption machine control and median deposit etc. technology, reduce prediction circuit
Area, and improve the working frequency of prediction circuit.
2nd, the implementation of the DC prediction algorithms proposed by the present invention applied to coding and decoding video, passes through the multiplexing of adder
Module, reduce computing repeatedly for data, logic gate number is reduced to 8762 from 12970, effectively reduce the occupancy face of circuit
Product, so as to reduce the design cost of whole circuit.
3rd, the input proposed by the present invention that adder is controlled by state machine, is answered adder so as to reach
With, while reduce redundant computation, and median cleverly is deposited using register, in original work period from 3 week
Phase drops to 2 cycles, and DC prediction algorithms are one kind of coding and decoding video prediction algorithm, because the speed of DC prediction algorithms circuit work
Degree improves 33%, so this will cause the speed of whole coding and decoding video to be obviously improved, and then improves work
Frequency, so as to improve the speed of coding and decoding video.
4th, the implementation of the DC prediction algorithms proposed by the present invention applied to coding and decoding video, has simplified circuit structure,
So that most fast working frequency reaches 500MHz, compared with circuit before, speed improves 25%, according to the work of circuit frequency
Rate, the real-time coding speed of video sequence is lifted to 7680 × 4320@30fps from 4096 × 2160@30fps.
5th, the DC prediction algorithm circuit engineerings after hardware optimization proposed by the present invention, more accurately follow DC prediction algorithms
Operation principle.When carrying out the calculating of Dcvalue values, it is contemplated that hardware circuit is in shifting process, it is possible that rounding up
Situation, adding N calculating to take into account, effectively overcome and all cast out the error predictions of appearance, the point for having pixel originally
Prediction into without pixel.The present invention compare before design, to algorithm realize it is more perfect, more accurate.
Brief description of the drawings
Fig. 1 is DC prediction modules of the present invention;
Fig. 2 is that DC of the present invention predicts hardware structure circuit;
Fig. 3 is the middle-value calculating circuit of DC prediction algorithms of the present invention;
Fig. 4 is the predictor calculation circuit of DC prediction algorithms of the present invention;
Fig. 5 is state machine circuit of the present invention;
Fig. 6 is optimum results experimental data comparison diagram of the present invention.
Embodiment
In the present embodiment, a kind of DC prediction circuits applied to coding and decoding video are used for any one in 2 × N cycle interior predictions
Pixel value in individual image in 4N × 4N regions, note 4N × 4N regions are a prediction block
Then prediction block PU top a line reference pixel reconstructed value is designated as A=[R0,-1,R1,-1,…,R4N-1,-1], prediction block PU left
One row reference pixel reconstructed value is designated as L=[R-1,0,R-1,1,…,R-1,4N-1];The note cycle is T, then TmRepresent m-th of cycle;
In specific implementation, N=1, then the pixel value in 2 cycle interior prediction any one image in 4 × 4 regions, note 4 ×
4 regions are a prediction blockAs shown in figure 1, then prediction block PU top a line reference image
Plain reconstructed value is designated as A=[R0,-1,R1,-1,R2,-1,R3-1], the prediction block PU row reference pixel reconstructed value of left one is designated as L=
[R-1,0,R-1,1,R-1,2,R-1,3];TmM-th of cycle is represented, then T1Represent the 1st cycle;Initialize n=1, m=1;
As shown in Fig. 2 DC prediction circuits include:Adder Multiplexing module, median register module, state self-adaption machine
Control module;
Adder Multiplexing module is in the 1st cycle T1It is interior that top a line reference pixel reconstructed value A and the row of left one are referred to
Pixel reconstructed value L carries out the 1st calculating, in obtaining 3 accumulated values SUM_0, SUM_4 and SUM_6 of the 1st calculating and passing to
Between value register module stored, and obtain the first sub-block of the prediction block PU when calculating for the 1st time
And directly export;
State self-adaption machine control module reads 3 accumulated value SUM_ of the 1st calculating from middle value register module
0th, SUM_4 and SUM_6 and adder Multiplexing module is passed to;
Adder Multiplexing module is in the 2nd cycle T2Interior three accumulated values SUM_0, SUM_4 and SUM_ to the 1st calculating
6 and top a line reference pixel reconstructed value A carries out the 1st calculating, prediction block PU the second sub-block when obtaining calculating for the 1st time
PU″1=[P0,0,P1,0,P2,0,P3,0];Three accumulated values SUM_0, SUM_4 and SUM_6 and left one of 1st calculating are arranged
Reference pixel reconstructed value L is calculated, and obtains the 3rd sub-block PU " ' of the prediction block PU when calculating for the 1st time1=[P0,1,P0,2,
P0,3];
The first sub-block PU when being calculated by the 1st time1', the second sub-block PU "1And the 3rd sub-block PU " '1Form the 1st calculating
When prediction block PU1, due to N=1, so n value is 1, i.e., it is once that can obtain predicted value PU only to need to calculate1=PU.
Adder Multiplexing module includes:No.1 adder, No. two adders, No. three adders, No. four adders, No. five plus
Musical instruments used in a Buddhist or Taoist mass, No. six adders, No. seven adders, No. eight adders, No.1 shift unit, No. two shift units, No. three shift units, No. four shiftings
Position device, No. five shift units, No. six shift units, No. seven shift units, No. eight shift units, No. nine shift units;
The 1st time of adder Multiplexing module is calculated as:
As shown in figure 3, in the 1st cycle T1Interior, No.1 adder to No. four adders is to top a line reference pixel weight
The built-in value A and row reference pixel reconstructed value L of left one carry out the 1st time calculating, obtain the 1st time calculating four accumulated value SUM_0,
SUM_1、SUM_2、SUM_3;
Four accumulated values SUM_0, SUM_1, SUM_2, SUM_3 and constant " 4 " of the 1st calculating are entered by No. eight adders
Row calculates, and obtains the accumulated value SUM_7 of the 1st calculating;
No.1 shift unit carries out moving to right 3 bit manipulations to the accumulated value SUM_7 of the 1st calculating, obtains the 1st calculating average value
DCvalue;
No. two shift units carry out moving to left 1 bit manipulation to the average value DCvalue of the 1st calculating, obtain the 1st time in calculating
Between value DCvalue ';
No. six adders are counted to the average value DCvalue of median DCvalue ' and the 1st calculating of the 1st calculating
Calculate, obtain the accumulated value SUM_5 of the 1st calculating;
No. seven adders are calculated the accumulated value SUM_5 and constant " 2 " of the 1st calculating, obtain the tired of the 1st calculating
Value added SUM_6;
No. five adders are calculated the median DCvalue ' and constant " 2 " of the 1st calculating, obtain the 1st calculating
Accumulated value SUM_4;
The first sub-block PU using the average value DCvalue of the 1st calculating as the 1st calculating1' in all elements value;
In the 2nd cycle T2It is interior, reuse eight adders of adder Module, the 1st cycle T1The centre of interior calculating
Value deposit is used as the 2nd cycle T into distributor module using it2Interior input;
As shown in figure 4, No.1 adder is counted to the SUM_4 of accumulated value SUM_0 and the 1st calculating of the 1st calculating
Calculate, obtained result recycles No. three shift units to carry out moving to right 2 bit manipulations, obtains the second sub-block PU " of the 1st calculating1In
First element value P0,0;
No. two adders are to the element value R in a line reference pixel reconstructed value A of top1,-1With the accumulated value of the 1st calculating
SUM_6 is calculated, and obtained result recycles No. four shift units to carry out moving to right 2 bit manipulations, obtains the second son of the 1st calculating
Block PU "1In second element value P1,0;
No. three adders are to the element value R in a line reference pixel reconstructed value A of top2,-1With the accumulated value of the 1st calculating
SUM_6 is calculated, and obtained result recycles No. five shift units to carry out moving to right 2 bit manipulations, obtains the second sub-block PU "1In
3rd element value P2,0;
No. four adders are to the element value R in a line reference pixel reconstructed value A of top3,-1With the accumulated value of the 1st calculating
SUM_6 is calculated, and obtained result recycles No. six shift units to carry out moving to right 2 bit manipulations, obtains the second son of the 1st calculating
Block PU "1In the 4th element value P3,0;
No. five adders are to the element value R in the row reference pixel reconstructed value L of left one-1,1With the accumulated value of the 1st calculating
SUM_6 is calculated, and obtained result recycles No. seven shift units to carry out moving to right 2 bit manipulations, obtains the 3rd son of the 1st calculating
Block PU " '1In first element value P0,1;
No. six adders are to the element value R in the row reference pixel reconstructed value L of left one-1,2With the accumulated value of the 1st calculating
SUM_6 is calculated, and obtained result recycles No. eight shift units to carry out moving to right 2 bit manipulations, obtains the 3rd son of the 1st calculating
Block PU " '1In second element value P0,2;
No. seven adders are to the element value R in the row reference pixel reconstructed value L of left one-1,3With the accumulated value of the 1st calculating
SUM_6 is calculated, and obtained result recycles No. nine shift units to carry out moving to right 2 bit manipulations, obtains the 3rd son of the 1st calculating
Block PU " '1In the 3rd element value P0,3;
As shown in figure 5, state self-adaption machine control module includes two states:Calculate intermediate value state M_value and pre-
Pixel state of value P_value is surveyed, two states are unconditionally to redirect automatically.
In this example, as shown in table 1, reference pixel value is by the row reference pixel reconstructed value of top one and the row reference image of left one
Plain reconstructed value is formed;
The reference pixel value of table 1
Top reference position | R0,-1 | R1,-1 | R2,-1 | R3,-1 |
Pixel value | 10 | 10 | 10 | 11 |
Left reference position | R-1,0 | R-1,1 | R-1,2 | R-1,3 |
Pixel value | 10 | 10 | 10 | 11 |
In the present embodiment, a kind of DC Forecasting Methodologies applied to coding and decoding video are to carry out as follows:
Step 1, initialization n=1, m=1;
Step 2, in the 1st cycle T1It is interior, obtain the accumulated value SUM_0 of the 1st calculating using formula (1):
SUM_0=R0,-1+R-1,0=20 (1)
Step 3, in the 1st cycle T1It is interior, the average value DCvalue of the 1st calculating is obtained using formula (2), and with average
First sub-blocks of the value DCvalue as the 1st calculatingIn all elements value;
Step 4, in the 1st cycle T1It is interior, obtain the accumulated value Sum_4 of the 1st calculating using formula (3):
Sum_4=2 × DCvalue+2=22 (3)
Step 5, in the 1st cycle T1It is interior, obtain the accumulated value Sum_6 of the 1st calculating using formula (4):
Sum_6=3 × DCvalue+2=32 (4)
Step 6, in the 2nd cycle T2It is interior, obtain the second sub-block PU " of the 1st calculating using formula (5)1=[P0,0,P1,0,
P2,0,P3,0] in first element value P0,0:
P0,0=(SUM_0+SUM_4) ÷ 4=10 (5)
Step 7, in the 2nd cycle T2It is interior, obtain the second sub-block PU " of the 1st calculating using formula (6)1In second member
Plain value P1,0:
P1,0=(R1,-1+ SUM_6) ÷ 4=10 (6)
Step 8, in the 2nd cycle T2It is interior, obtain the second sub-block PU " of the 1st calculating using formula (7)1In the 3rd member
Plain value P2,0:
P2,0=(R2,-1+ SUM_6) ÷ 4=10 (7)
Step 9, in the 2nd cycle T2It is interior, obtain the second sub-block PU " of the 1st calculating using formula (8)1In the 4th member
Plain value P3,0:
P3,0=(R3,-1+ SUM_6) ÷ 4=10 (8)
Step 10, in the 2nd cycle T2It is interior, obtain the 3rd sub-block PU " ' of the 1st calculating using formula (9)1In first
Element value P0,1:
P0,1=(R-1,1+ SUM_6) ÷ 4=10 (9)
Step 11, in the 2nd cycle T2It is interior, obtain the 3rd sub-block PU " ' of the 1st calculating using formula (10)1In second
Individual element value P0,2:
P0,2=(R-1,2+ SUM_6) ÷ 4=10 (10)
Step 12, in the 2nd cycle T2It is interior, obtain the 3rd sub-block PU " ' of the 1st calculating using formula (11)1In the 3rd
Individual element value P0,3:
P0,3=(R-1,3+ SUM_6) ÷ 4=10 (11)
So as to obtain prediction block the second sub-blocks of PU PU " when calculating for the 1st time1=[P0,0,P1,0,P2,0,P3,0] and the 3rd son
Block PU " '1=[P0,1,P0,2,P0,3];
Step 13, the second sub-block PU " by the 1st calculating1With the 3rd sub-block PU " '1Form prediction block when calculating the 1st time
PU1;
Step 14, by 1≤n≤N;1≤m≤2 × N, which knows, only to be needed to calculate once, so as to obtain prediction block PU value
Such as table 2;
The predicted pixel values of table 2
Predicted position | P0,0 | P1,0 | P2,0 | P3,0 |
Pixel value | 10 | 10 | 10 | 10 |
Predicted position | P0,1 | P1,1 | P2,1 | P3,1 |
Pixel value | 10 | 10 | 10 | 10 |
Predicted position | P0,2 | P1,2 | P2,2 | P3,2 |
Pixel value | 10 | 10 | 10 | 10 |
Predicted position | P0,3 | P1,3 | P2,3 | P3,3 |
Pixel value | 10 | 10 | 10 | 10 |
VLSI designs are carried out to DC prediction algorithms for 4 × 4 prediction block, and are retouched using Verilog HDL language
State, emulated and integrated by Synopsys softwares, the comprehensive circuit that obtains most works frequently soon under SMIC 0.18um techniques
Rate is 500MHz, and the area of circuit work is 8762gate, and the algorithm cycle is reduced to 2 cycles.The present invention and Zhou Wei, Huang Xiaodong
Et al. deliver " EFFICIENT INTRA PREDICTION VLSI in 2013PCS (Picture Coding Symposium)
Circuit is compared described in ARCHITECTURE FOR HEVC STANDARD ", and working frequency improves 40%, and gate number reduces
40%, the algorithm cycle reduces 33%, working frequency, logic gate number, the comparison diagram in algorithm cycle respectively shown in Fig. 6, if
Expand on the platform of whole coding and decoding video, effect of optimization becomes apparent from.
Claims (3)
1. a kind of DC prediction circuits applied to coding and decoding video, it is characterized in that, in any one figure of 2 × N cycles interior prediction
Pixel value as in 4N × 4N regions, it is a prediction block to remember 4N × 4N regions
Then top a line reference pixel reconstructed value of the prediction block PU is designated as A=[R0,-1,R1,-1,…,R4N-1,-1], the prediction block
The PU row reference pixel reconstructed value of left one is designated as L=[R-1,0,R-1,1,…,R-1,4N-1];It is T to remember the cycle, then TmRepresent
M-th of cycle;Initialize n=1, m=1;
The DC prediction circuits include:Adder Multiplexing module, median register module, state self-adaption machine control module;
The adder Multiplexing module is in m-th of cycle TmIt is interior that top a line reference pixel reconstructed value A and the row of left one are joined
Examine pixel reconstructed value L and carry out n-th calculating, obtain 3 accumulated values SUM_0, SUM_4 and SUM_6 of n-th calculating and pass to
The median register module is stored, and obtains the first sub-block of the prediction block PU when n-th calculatesAnd directly export;
The state self-adaption machine control module reads 3 that the n-th calculates from the median register module and tired out
Value added SUM_0, SUM_4 and SUM_6 simultaneously pass to the adder Multiplexing module;
The adder Multiplexing module is in the m+1 cycle Tm+1Interior three accumulated values SUM_0, SUM_ calculated the n-th
4 and SUM_6 and top a line reference pixel reconstructed value A carries out n-th calculating, obtains prediction when n-th calculates
Block PU the second sub-block PUn"=[P0,0,P1,0,…,P4n-1,0];Three accumulated values SUM_0, the SUM_4 calculated the n-th
Calculated with SUM_6 and the row reference pixel reconstructed value L of the left one, obtain the prediction block PU when n-th calculates
The 3rd sub-block PUn" '=[P0,1,P0,2,…,P0,4n-1];
The first sub-block PU when being calculated by the n-thn', the second sub-block PUn" and the 3rd sub-block PUn" ' form n-th calculates
When prediction block PUn, n+1 is assigned to n, after m+2 is assigned into m, the n-th for repeating the adder Multiplexing module calculates, directly
Untill n=N and m=2 × N, so as to obtain prediction block PU.
2. the DC prediction circuits according to claim 1 applied to coding and decoding video, it is characterized in that, the adder multiplexing
Module includes:No.1 adder, No. two adders, No. three adders, No. four adders, No. five adders, No. six adders, seven
Number adder, No. eight adders, No.1 shift unit, No. two shift units, No. three shift units, No. four shift units, No. five shift units, six
Number shift unit, No. seven shift units, No. eight shift units, No. nine shift units;
The n-th of the adder Multiplexing module is calculated as:
In m-th of cycle TmIt is interior, the No.1 adder to No. four adders to top a line reference pixel reconstructed value A and
The row reference pixel reconstructed value L of left one carry out n-th calculating, obtain n-th calculating four accumulated value SUM_0, SUM_1,
SUM_2、SUM_3;
Four accumulated values SUM_0, SUM_1, SUM_2, the SUM_3 and constant calculated by No. eight adders the n-th
" 4 " are calculated, and obtain the accumulated value SUM_7 of n-th calculating;
The No.1 shift unit carries out moving to right 3 bit manipulations to the accumulated value SUM_7 that the n-th calculates, and it is flat to obtain n-th calculating
Average DCvalue;
No. two shift units carry out moving to left 1 bit manipulation to the average value DCvalue that the n-th calculates, and obtain n-th calculating
Median DCvalue ';
The average value DCvalue that the median DCvalue ' and n-th that No. six adders are calculated the n-th are calculated enters
Row calculates, and obtains the accumulated value SUM_5 of n-th calculating;
The accumulated value SUM_5 and constant " 2 " that No. seven adders are calculated the n-th are calculated, and obtain n-th calculating
Accumulated value SUM_6;
The median DCvalue ' and constant " 2 " that No. five adders are calculated the n-th are calculated, and obtain n-th
The accumulated value SUM_4 of calculating;
The first sub-block PU that the average value DCvalue calculated using the n-th calculates as the n-thn' in all elements
Value;
In the m+1 cycle Tm+1Interior, the accumulated value SUM_0 and n-th that the No.1 adder is calculated the n-th are calculated
SUM_4 carry out n-th calculating, obtained result recycles No. three shift units to carry out moving to right 2 bit manipulations, obtains n-th
The the second sub-block PU calculatedn" in first element value Pn-1,n-1;
No. two adders are to the element value R in top a line reference pixel reconstructed value An,-1Calculated with n-th cumulative
Value SUM_6 is calculated, and obtained result recycles No. four shift units to carry out moving to right 2 bit manipulations, obtains n-th calculating
Second sub-block PUn" in second element value Pn,n-1;
No. three adders are to the element value R in top a line reference pixel reconstructed value An+1,-1Calculated with n-th tired
Value added SUM_6 is calculated, and obtained result recycles No. five shift units to carry out moving to right 2 bit manipulations, obtains the second sub-block
PUn" in the 3rd element value Pn+1,n-1;
No. four adders are to the element value R in top a line reference pixel reconstructed value An+2,-1Calculated with n-th tired
Value added SUM_6 is calculated, and obtained result recycles No. six shift units to carry out moving to right 2 bit manipulations, obtains n-th calculating
The second sub-block PUn" in the 4th element value Pn+2,n-1;
No. five adders are to the element value R in the row reference pixel reconstructed value L of left one-1,nCalculated with n-th cumulative
Value SUM_6 is calculated, and obtained result recycles No. seven shift units to carry out moving to right 2 bit manipulations, obtains n-th calculating
3rd sub-block PUn" ' in first element value Pn-1,n;
No. six adders are to the element value R in the row reference pixel reconstructed value L of left one-1,n+1Calculated with n-th tired
Value added SUM_6 is calculated, and obtained result recycles No. eight shift units to carry out moving to right 2 bit manipulations, obtains n-th calculating
The 3rd sub-block PUn" ' in second element value Pn-1,n+1;
No. seven adders are to the element value R in the row reference pixel reconstructed value L of left one-1,n+2Calculated with n-th tired
Value added SUM_6 is calculated, and obtained result recycles No. nine shift units to carry out moving to right 2 bit manipulations, obtains n-th calculating
The 3rd sub-block PUn" ' in the 3rd element value Pn-1,n+2。
A kind of 3. DC Forecasting Methodologies applied to coding and decoding video, it is characterized in that in any one figure of 2 × N cycles interior prediction
Pixel value as in 4N × 4N regions, it is a prediction block to remember 4N × 4N regions
Then top a line reference pixel reconstructed value of the prediction block PU is designated as A=[R0,-1,R1,-1,…,R4N-1,-1], the prediction block
The PU row reference pixel reconstructed value of left one is designated as L=[R-1,0,R-1,1,…,R-1,4N-1];It is T to remember the cycle, then TmRepresent
M-th of cycle;N ∈ (1,2,4,8,16);1≤n≤N;1≤m≤2×N;
The Forecasting Methodology is to carry out as follows:
Step 1, initialization n=1, m=1;
Step 2, in m-th of cycle TmIt is interior, obtain the accumulated value SUM_0 of n-th calculating using formula (1):
SUM_0=Rn-1,-1+R-1,n-1 (1)
Step 3, in m-th of cycle TmIt is interior, the average value DCvalue of n-th calculating is obtained using formula (2), and with the average value
The first sub-block that DCvalue calculates as the n-thIn all elements value;
<mrow>
<mi>D</mi>
<mi>C</mi>
<mi>v</mi>
<mi>a</mi>
<mi>l</mi>
<mi>u</mi>
<mi>e</mi>
<mo>=</mo>
<mrow>
<mo>(</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>x</mi>
<mo>=</mo>
<mn>4</mn>
<mi>n</mi>
<mo>-</mo>
<mn>4</mn>
</mrow>
<mrow>
<mn>4</mn>
<mi>n</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</munderover>
<msub>
<mi>R</mi>
<mrow>
<mi>x</mi>
<mo>,</mo>
<mo>-</mo>
<mn>1</mn>
</mrow>
</msub>
<mo>+</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>y</mi>
<mo>=</mo>
<mn>4</mn>
<mi>n</mi>
<mo>-</mo>
<mn>4</mn>
</mrow>
<mrow>
<mn>4</mn>
<mi>n</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</munderover>
<msub>
<mi>R</mi>
<mrow>
<mo>-</mo>
<mn>1</mn>
<mo>,</mo>
<mi>y</mi>
</mrow>
</msub>
<mo>+</mo>
<mn>4</mn>
<mi>N</mi>
<mo>)</mo>
</mrow>
<mo>></mo>
<mo>></mo>
<mrow>
<mo>(</mo>
<msubsup>
<mi>log</mi>
<mn>2</mn>
<mrow>
<mn>4</mn>
<mi>N</mi>
</mrow>
</msubsup>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>2</mn>
<mo>)</mo>
</mrow>
</mrow>
Step 4, in m-th of cycle TmIt is interior, obtain the accumulated value Sum_4 of n-th calculating using formula (3):
Sum_4=2 × DCvalue+2 (3)
Step 5, in m-th of cycle TmIt is interior, obtain the accumulated value Sum_6 of n-th calculating using formula (4):
Sum_6=3 × DCvalue+2 (4)
Step 6, in the m+1 cycle Tm+1It is interior, obtain the second sub-block PU of n-th calculating using formula (5)n"=[P0,0,
P1,0,…,P4n-1,0] in first element value Pn-1,n-1:
Pn-1,n-1=(SUM_0+SUM_4) ÷ 4 (5)
Step 7, in the m+1 cycle Tm+1It is interior, obtain the second sub-block PU of n-th calculating using formula (6)n" in second member
Plain value Pn,n-1:
Pn,n-1=(Rn,-1+SUM_6)÷4 (6)
Step 8, in the m+1 cycle Tm+1It is interior, obtain the second sub-block PU of n-th calculating using formula (7)n" in the 3rd member
Plain value Pn+1,n-1:
Pn+1,n-1=(Rn+1,-1+SUM_6)÷4 (7)
Step 9, in the m+1 cycle Tm+1It is interior, obtain the second sub-block PU of n-th calculating using formula (8)n" in the 4th member
Plain value Pn+2,n-1:
Pn+2,n-1=(Rn+2,-1+SUM_6)÷4 (8)
Step 10, in the m+1 cycle Tm+1It is interior, obtain the 3rd sub-block PU of n-th calculating using formula (9)n" ' in first
Element value Pn-1,n:
Pn-1,n=(R-1,n+SUM_6)÷4 (9)
Step 11, in the m+1 cycle Tm+1It is interior, obtain the 3rd sub-block PU of n-th calculating using formula (10)n" ' in second
Element value Pn-1,n+1:
Pn-1,n+1=(R-1,n+1+SUM_6)÷4 (10)
Step 12, in the m+1 cycle Tm+1It is interior, obtain the 3rd sub-block PU of n-th calculating using formula (11)n" ' in the 3rd
Element value Pn-1,n+2:
Pn-1,n+2=(R-1,n+2+SUM_6)÷4 (11)
So as to obtain the second sub-blocks of prediction block PU PU when n-th calculatesn"=[P0,0,P1,0,…,P4n-1,0] and the 3rd son
Block PUn" '=[P0,1,P0,2,…,P0,4n-1];
Step 13, the first sub-block PU when being calculated by the n-th1', the second sub-block PUn" and the 3rd sub-block PUn" ' form n-th
Prediction block PU during secondary calculatingn;
Step 14, n+1 being assigned to n, after m+2 is assigned into m, repeat step 2 performs, untill n=N and m=2 × N, from
And obtain prediction block PU.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610471667.9A CN105898334B (en) | 2016-06-22 | 2016-06-22 | A kind of DC prediction circuits and its method applied to coding and decoding video |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610471667.9A CN105898334B (en) | 2016-06-22 | 2016-06-22 | A kind of DC prediction circuits and its method applied to coding and decoding video |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105898334A CN105898334A (en) | 2016-08-24 |
CN105898334B true CN105898334B (en) | 2017-12-05 |
Family
ID=56718808
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610471667.9A Active CN105898334B (en) | 2016-06-22 | 2016-06-22 | A kind of DC prediction circuits and its method applied to coding and decoding video |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105898334B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106993192B (en) * | 2017-04-14 | 2019-07-19 | 合肥工业大学 | The angle prediction circuit and its prediction technique of intra prediction in coding and decoding video |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101534373A (en) * | 2009-04-24 | 2009-09-16 | 北京空间机电研究所 | Remote sensing image near-lossless compression hardware realization method based on improved JPEG-LS algorithm |
CN104038765A (en) * | 2014-06-26 | 2014-09-10 | 天津大学 | Rapid and efficient damage-free image compression method oriented to hardware achievement |
CN104683800A (en) * | 2015-02-11 | 2015-06-03 | 广州柯维新数码科技有限公司 | AVS-based methods for parallel quantization and inverse quantization |
CN105208394A (en) * | 2015-09-21 | 2015-12-30 | 北京集创北方科技有限公司 | Real-time digital image compression prediction method and system |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN100413344C (en) * | 2006-10-20 | 2008-08-20 | 清华大学 | Method for realizing high-parallel frame predicator |
CN101115207B (en) * | 2007-08-30 | 2010-07-21 | 上海交通大学 | Method and device for implementing interframe forecast based on relativity between future positions |
CN102131081A (en) * | 2010-01-13 | 2011-07-20 | 华为技术有限公司 | Dimension-mixed coding/decoding method and device |
KR101307257B1 (en) * | 2012-06-28 | 2013-09-12 | 숭실대학교산학협력단 | Apparatus for video intra prediction |
CN105635731B (en) * | 2016-01-07 | 2018-11-20 | 西安电子科技大学 | The inter-frame predicated reference point preprocess method of efficient video coding |
-
2016
- 2016-06-22 CN CN201610471667.9A patent/CN105898334B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101534373A (en) * | 2009-04-24 | 2009-09-16 | 北京空间机电研究所 | Remote sensing image near-lossless compression hardware realization method based on improved JPEG-LS algorithm |
CN104038765A (en) * | 2014-06-26 | 2014-09-10 | 天津大学 | Rapid and efficient damage-free image compression method oriented to hardware achievement |
CN104683800A (en) * | 2015-02-11 | 2015-06-03 | 广州柯维新数码科技有限公司 | AVS-based methods for parallel quantization and inverse quantization |
CN105208394A (en) * | 2015-09-21 | 2015-12-30 | 北京集创北方科技有限公司 | Real-time digital image compression prediction method and system |
Also Published As
Publication number | Publication date |
---|---|
CN105898334A (en) | 2016-08-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN101222646B (en) | Infra-frame prediction device and method suitable for AVS encoding | |
CN108733348A (en) | The method for merging vector multiplier and carrying out operation using it | |
CN105183425B (en) | A kind of fixation bit wide multiplier with high-precision low complex degree characteristic | |
CN101848311B (en) | JPEG2000 EBCOT encoder based on Avalon bus | |
Kim et al. | Mosaic-CNN: A combined two-step zero prediction approach to trade off accuracy and computation energy in convolutional neural networks | |
CN102148990B (en) | Device and method for predicting motion vector | |
Wahid et al. | Efficient hardware implementation of 8× 8 integer cosine transforms for multiple video codecs | |
Azgin et al. | A computation and energy reduction technique for HEVC intra prediction | |
CN104244010A (en) | Method for improving digital signal conversion performance and digital signal conversion method and device | |
CN105898334B (en) | A kind of DC prediction circuits and its method applied to coding and decoding video | |
Kalali et al. | An approximate HEVC intra angular prediction hardware | |
CN100469146C (en) | Video image motion compensator | |
CN103188487B (en) | Convolution method in video image and video image processing system | |
CN106060565B (en) | A kind of Planar prediction circuit and Planar prediction technique applied to coding and decoding video | |
Nagaraju et al. | Design and implementation of an efficient mixed parallel-pipeline SAD architecture for HEVC motion estimation | |
CN113360131A (en) | Logarithm approximate multiplication accumulator for convolutional neural network accelerator | |
CN102055980B (en) | Intra-frame predicting circuit for video coder and realizing method thereof | |
CN110737869A (en) | DCT/IDCT multiplier circuit optimization method and application | |
Cobrnic et al. | An Area Efficient and Reusable HEVC 1D-DCT Hardware Accelerator | |
Yagain et al. | High Speed ASIC Design of DCT for Image Compression | |
Mohanty et al. | Area‐delay‐power‐efficient architecture for folded two‐dimensional discrete wavelet transform by multiple lifting computation | |
Deng et al. | The FPGA design of JPEG-LS image lossless decompression IP core | |
Jou et al. | A low-cost gray prediction search chip for motion estimation | |
CN108271025B (en) | The coding circuit of depth modelling mode and its coding method in 3D coding and decoding video based on boundary gradient | |
Taşdizen et al. | High performance hardware architectures for a hexagon-based motion estimation algorithm |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
TR01 | Transfer of patent right | ||
TR01 | Transfer of patent right |
Effective date of registration: 20201225 Address after: 245000 No.50, Meilin Avenue, Huangshan Economic Development Zone, Huangshan City, Anhui Province Patentee after: Huangshan Development Investment Group Co.,Ltd. Address before: Tunxi road in Baohe District of Hefei city of Anhui Province, No. 193 230009 Patentee before: Hefei University of Technology |