CN104954793A - GOP-level QP-Offset setting method - Google Patents

GOP-level QP-Offset setting method Download PDF

Info

Publication number
CN104954793A
CN104954793A CN201510341367.4A CN201510341367A CN104954793A CN 104954793 A CN104954793 A CN 104954793A CN 201510341367 A CN201510341367 A CN 201510341367A CN 104954793 A CN104954793 A CN 104954793A
Authority
CN
China
Prior art keywords
gop
level
offset
big
frame
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.)
Granted
Application number
CN201510341367.4A
Other languages
Chinese (zh)
Other versions
CN104954793B (en
Inventor
朱策
徐丽英
周益民
王勇
钟敏
高艳博
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
University of Electronic Science and Technology of China
Original Assignee
University of Electronic Science and Technology of China
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by University of Electronic Science and Technology of China filed Critical University of Electronic Science and Technology of China
Priority to CN201510341367.4A priority Critical patent/CN104954793B/en
Publication of CN104954793A publication Critical patent/CN104954793A/en
Application granted granted Critical
Publication of CN104954793B publication Critical patent/CN104954793B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Abstract

The invention provides a GOP-level QP-Offset setting method, which belongs to the technical field of video processing, and particularly relates to a method for setting quantization parameters under all supporting random access (RA) structure conditions. The GOP-level QP-Offset setting method comprises the following steps of through analyzing the reference relationship between GOP under the RA structure, which has the transferred meaning of the effect strength of the current GOP on the following GOP, describing different importance of the GOP through effect strength attenuation factors of the GOP, and adopting the different importance of the GOP as the basis of GOP-level QP-Offset distribution. According to the GOP-level QP-Offset setting method, the time domain redundant information can be further eliminated on the basis of an original frame-level layering QP-Offset technology. The experiments on two platforms of HEVC and AVS2 show that without increasing time complexity, the bit is further saved, and the encoding performance is improved.

Description

A kind of QP-Offset method to set up of GOP level
Technical field
The invention belongs to technical field of video processing, particularly the quantization parameter method to set up of GOP level under all support random accesss (Random Access, RA) structure.
Background technology
In recent years, people are to high definition, and the demand of ultra high-definition video is more and more vigorous, therefore the demand of higher compression efficiency also gets more and more.The encoder of current main flow still adopts traditional hybrid encoding frame, mainly comprises prediction, conversion, quantification, filtering, a few part composition of entropy code.
Video signal process, by some time continuous print image construction, can become several image sets (Group of Pictures GOP) to carry out by video sequence in coding.Random access (Random-access, RA) coding structure is time domain hierarchy (Hierarchical B structure, HBP), be a GOP with 8 frames under normal circumstances, and periodically insert a random access frame (Clean Random Access, CRA).By decoding order, two I frames be spaced apart a BIG-GOP, multiple GOP can be comprised in a usual BIG-GOP.
Quantification can realize conversion coefficient many-one and map, and Video coding quantizes gain of parameter compression performance by optimizing, but also introduces quantizing distortion simultaneously.Whether the setting of quantization parameter (Quantization Parameter, QP) and rate-distortion optimization Lagrange coefficient Lambda is proper will be directly connected to the quality of coding efficiency.The QP value excursion of HEVC definition is [0-51], the QP value excursion of AVS-2 definition is [0-63], the calculating of Lambda value is frame type according to each Slice, the degree of depth, QP value, QP factor value and calculating obtains, can be understood as under the coding sight determined, when QP value is to timing, Lambda value also will be calculated.
For eliminating the temporal redundancy information of interframe, adopt the QP migration technology (i.e. QP-Offset) of frame level to promote coding efficiency.Current encoder is with basic QP value (QP base) add frame level complete the QP value setting of present frame, as shown in formula (1):
QP i = QP Base + QP Offset Pic ( l ) - - - ( 1 )
Wherein, l represents the level that the i-th frame is residing in GOP.
Based on considering of code efficiency and picture quality, under RA structure, I-frame uses basic QP value, and be the given different QP-Offset value of various level B/P frame in a GOP, as shown in Figure 1, be that the GOP of 8 frame 4 layers has formulated side-play amount simultaneously be respectively 1 (layer 0), 2 (layer 1), 3 (layer 2), 4 (layer 3).
RA coding needs periodically to insert random access frame (Clean Random Access, CRA) i.e. I-two field picture, random access point is provided with this, and provide high-quality reference frame for follow-up GOP, meanwhile, the I-frame of new insertion will intercept the referring-to relation of current GOP and previous GOP on very large probability, therefore its importance of the difference of GOP present position is also different.But the frame stage layered QP-Offset technology under RA coding structure only considered the frame-to-frame correlation of GOP inside, not yet considers the reference correlation between GOP.
Summary of the invention
For promoting code efficiency, the present invention proposes and a kind of control method is set for the GOP level QP-Offset under RA structure, supporting all configurations based on RA structure.
The present invention specifically adopts following technical scheme:
A QP-Offset method to set up for GOP level, its flow process as shown in Figure 2, specifically comprises the following steps:
Step 1. initialization;
Step 1-1. define a BIG-GOP comprise n GOP and start from scratch numbering to n-1;
The QP-Offset of definition GOP level is expressed as frame level QP value then in formula (1) calculates and will be revised as shown in formula (2):
QP i = QP Base + QP Offset Base ( k ) + QP Offset Pic ( l ) - - - ( 2 )
Wherein, QP irefer to the quantization parameter QP of the i-th two field picture under RA coding structure, k represents the Position Number of GOP in its BIG-GOP belonging to frame i, and k ∈ [0, n-1], l represent hierarchical location residing in the GOP of frame i belonging to it;
Step 1-2. is in the inside of single GOP, and inter-reference relation is all determined and fixed, and the present invention does not relate to the referring-to relation of GOP inside;
For the referring-to relation across GOP, the present invention is defined as directly reference and one-level reference from low to high successively with reference to relation, indirect reference and secondary reference, three grades of references, level Four reference, Pyatyi reference ... reference level other definition by that analogy, across the direct reference of GOP and indirect reference relation as shown in Figure 3, wherein solid line represents direct reference, dotted line represents indirect reference: frame 8 is reference frame POC 0 directly, frame POC 9 is reference frame POC 8 directly, just mean frame POC 9 through frame POC 8 indirect reference frame POC 0, when directly reference is determined, indirect reference relation also will be determined,
Step 1-3. this method definition weight factor α affects intensity, 0< α <1 with what weigh direct reference, and getting α value in this method is without loss of generality 1/2; Along with the increase of referring-to relation progression τ, the affecting intensity and will decay of reference, referring-to relation progression is for affecting intensity w (τ) definition as shown in formula (3) during τ:
w(τ)=α τ,τ=1,2,..,m (3)
Wherein, m is the highest with reference to relative degree number;
For needing RA coding structure to be processed, each GOP's in one of them BIG-GOP arrange by method described in step 2 to step 4;
Step 2. adds up referring-to relation number ∑ N (τ) of each GOP;
Be numbered all frames in the GOP of k and be designated as N by the sum of the τ level reference of the frame of other GOP k(τ), the such as all frames of this GOP are designated as N by the frame one-level of other GOP with reference to sum k(1), secondary is designated as N with reference to sum k(2), three grades are designated as N with reference to sum k(3);
In the statistic processes of referring-to relation number, if a certain frame A in this GOP respectively by two different with reference to progression by a certain frame B reference of other GOP, then when adding up the referring-to relation number of this GOP, the minimum referring-to relation of statistics frame A and frame B and do not record the referring-to relation of other ranks, for Fig. 4: frame POC 24 is reference frame POC 8 directly, with time frame POC 24 through frame POC 16 indirect reference frame POC 8, if now add up the referring-to relation number of GOP-1, " frame POC 24 is reference frame POC 8 directly " should be counted corresponding N k(1), and not " frame POC 24 through frame POC 16 indirect reference frame POC 8 " is not added up into N k(2);
The GOP that step 3. is numbered k always affects intensity W kas shown in formula (4):
W k = &Sigma; &tau; = 1 m w k ( &tau; ) &CenterDot; N k ( &tau; ) = &Sigma; &tau; = 1 m &alpha; &tau; &CenterDot; N k ( &tau; ) - - - ( 4 )
Definition GOP affects strength retrogression's Graded factor Δ W ias shown in formula (5):
ΔW i=|W i-W i-1|,i=2,3,……,n-2,n-1 (5)
Step 4. the present invention establishes the mapping relations affected between strength retrogression's Graded factor and GOP level QP-Offset of GOP, affects strength retrogression's Graded factor Δ W with GOP iportray the importance of GOP, the importance affecting the current GOP of the larger explanation of intensity is larger, and it is more that the larger explanation of the attenuation gradient factor affects remitted its fury, and importance is lower, therefore can in this, as the foundation of distributing GOP level QP-Offset; The importance of GOP presents the dull rule reduced, that is: as Δ W itime minimum, can think that current GOP affects intensity still very strong; As Δ W iwhen becoming large, illustrate that the importance of GOP weakens, therefore the QP-Offset distributing GOP level presents the rule increased progressively, concrete distribution method is as follows:
Step 4-1. is by Δ W inormalize to [0,1] interval and be designated as Δ W norm – i, as shown in formula (6):
ΔW norm-i=(ΔW i-ΔW 2)/(ΔW n-1-ΔW 2) (6)
Step 4-2. integrated encode quality and picture quality consider, and definition GOP level QP-Offset namely adjustable extent be [0,2];
The present invention does not consider that in BIG-GOP, GOP number is less than or equal to the situation of 2, does not make the amendment of GOP level QP-Offset in the present invention;
If the number of GOP equals 3 in a BIG-GOP, the adjustable extent of definition GOP level QP-Offset is [0,1]: directly the GOP level QP-Offset of definition the first two GOP is the GOP level QP-Offset of the zero, three GOP is 1;
If the number of GOP is greater than 4 in BIG-GOP, then perform step 4-3 to step 4-4;
If the number of GOP equals 4 in BIG-GOP, now GOP to affect intensity all less, for ensureing that coding quality does not make side-play amount to the first two GOP of BIG-GOP and changes, i.e. the first two GOP in BIG-GOP all be set to zero, the side-play amount of residue GOP it is as follows that rule is set:
(1) if Δ W norm – i≤ 0.1, be then encoded to the GOP's of i be 1.
(2) if Δ W norm – i>0.1, be then numbered the GOP's of i be set to 2;
Step 4-3. sets self-defined minimum ε, and its span is (0,0.1);
If the number of GOP is greater than 4 in BIG-GOP, then the GOP level QP-Offset that the GOP being numbered zero in GOP and BIG-GOP at I-frame place distributes is set to zero; If be numbered the Δ W of 3 i.e. the 4th GOP norm – i< ε, being then numbered 1 and 2 is second and third GOP's be set to zero, otherwise be set to 1; The side-play amount that residue GOP is corresponding arrange by the rule that arranges described in this step 4-4;
Arranging of the GOP level QP-Offset of GOP is remained regular as follows in this BIG-GOP of step 4-4.:
(1) if Δ W norm – i< ε, be then numbered the GOP's of i be set to 0;
(2) if ε≤Δ W norm – i≤ 0.1, be then numbered the GOP's of i be set to 1;
(3) if Δ W norm – i>0.1, be then numbered the GOP's of i be set to 2;
Step 4-5., for needing RA coding structure to be processed, can obtain as stated above to each GOP's in each BIG-GOP the QP-Offset realizing the GOP level of RA coding structure is thus arranged.
The invention has the beneficial effects as follows:
The present invention is by the fixing referring-to relation under analysis RA structure between GOP, directly can read configuration file to obtain, affect with GOP the different importance that strength retrogression's factor portrays GOP, original frame stage layered QP-Offset technical foundation can eliminate temporal redundancy information further; And known by the actual result on HEVC, AVS2 two experiment porchs, not increasing on the basis of time complexity, save bit further, complexity of saving time also promotes coding efficiency.
Accompanying drawing explanation
Fig. 1 is GOP internal reference relation;
Fig. 2 is the inventive method flow chart;
Fig. 3 is across GOP directly reference and indirect reference schematic diagram;
Fig. 4 is across GOP direct interact relation when in AVS2, GOP quantity is 4;
Fig. 5 affects intensity when GOP number is n in AVS2 in embodiment;
Fig. 6 is Δ W between GOP in AVS2 in embodiment idistribution;
Fig. 7 affects intensity when GOP number is n in HEVC in embodiment;
Fig. 8 is Δ W between GOP in HEVC in embodiment idistribution;
Fig. 9 is that the QP-Offset of GOP level under the general BIG-GOP size of AVS2 in embodiment is arranged;
Figure 10 is that the QP-Offset of GOP level under the general BIG-GOP size of HEVC in embodiment is arranged;
Figure 11 is GOP level QP-Offset control method test result under RA structure on AVS2 platform in embodiment;
Figure 12 is GOP level QP-Offset control method test result under RA structure on HEVC platform in embodiment.
Embodiment
Below in conjunction with drawings and Examples, the present invention is described further.
Embodiment
In order to the more clear QP-Offset method to set up representing the GOP level that this method proposes, table 1 gives setting procedure.
Table 1 GOP level QP-Offset method to set up
In the present embodiment, α value is empirical value 1/2; The present embodiment has added up the referring-to relation in BIG-GOP under general GOP number (3/4/6/8) between GOP, and extends to the situation that GOP number is n, as shown in Figure 5.In Fig. 5, Fig. 7, Δ Wi all obeys certain fixed attenuation rule, from last GOP (n-1), is one group with two GOP, and first GOP group amplitude is (Δ W n-1, Δ W n-2), amplitude presents the exponential decay of 1/2, as shown in formula (7); Namely the strength retrogression's Graded factor trend minimum 0, Fig. 6 that affects from GOP-(n-1) to GOP-1 lists Δ W when GOP number is 25 isituation.
&Delta;W n - s = &Delta;W n - 1 &CenterDot; ( 1 2 ) s - 1 2 ( s mod 2 = 1 ) &Delta;W n - 2 &CenterDot; ( 1 2 ) s - 2 2 ( s mod 2 = 0 ) , s = 1,2,3 , . . . . . . , n - 1 - - - ( 7 )
Affect intensity when Fig. 5 illustrates that in AVS2, in BIG-GOP, GOP number is n, Fig. 6 to illustrate in AVS2 Δ W between GOP idistribution.
Affect intensity when Fig. 7 illustrates that in HEVC, in BIG-GOP, GOP number is n, Fig. 8 to illustrate in HEVC Δ W between GOP idistribution.
Under the BIG-GOP size that current AVS2 coding techniques is general, Fig. 9 illustrates the QP-Offset facilities of various GOP level.
Under the BIG-GOP size that current HEVC coding techniques is general, Figure 10 illustrates the QP-Offset facilities of various GOP level.
Lead to survey condition according to AVS2, this method GOP level QP-Offset method to set up realizes on RD11.0, and under RA configuration, carry out logical survey according to CTC, Figure 11 gives test result.It should be noted that in logical survey, Refind QP technology is set to acquiescence and opens.As can see from Figure 11, under various resolution, this method can obtain gain in gain in various degree, particularly WVGA and 720p and exceed 1%, and total average BD-Rate gain reaches 0.82% in the Y direction, illustrates that this method performance boost is obvious.
Lead to survey condition according to HEVC, this method GOP level QP-Offset method to set up realizes on HM16.4, and under RA configuration, carry out logical survey according to CTC, Figure 12 gives test result.As can see from Figure 12, under various resolution, this method can obtain gain in various degree, particularly Class C D on F gain exceeded 1%, total average BD-Rate gain reaches 1.0% in the Y direction, illustrates that this method performance boost is obvious.
Weigh current GOP according to referring-to relation fixing between GOP under RA structure and intensity is affected on follow-up GOP, and then describe the different importance of GOP, for each GOP is assigned with different QP-Offset.This method can eliminate temporal redundancy information further, can save bit, improving performance, and BD-Rate gain is obvious.From time loss, this method saves part-time in coding.Obtain BD-rate gain 1.0% under the RA that this method is implemented to obtain BD-Rate gain 0.82%, HM16.4 under the RA of RD11.0 leads to survey condition leads to survey condition, performance boost is obvious.

Claims (2)

1. a QP-Offset method to set up for GOP level, specifically comprises the following steps:
Step 1. initialization;
Step 1-1. define a complete BIG-GOP comprise n GOP and start from scratch numbering to n-1;
The QP-Offset of definition GOP level is expressed as frame level QP value calculates as shown in formula (2):
QP i = QP Base + QP Offset Base ( k ) + QP Offset Pic ( l ) - - - ( 2 )
Wherein, QP irefer to the quantization parameter QP of the i-th two field picture under RA coding structure, k represents the position number of GOP in its BIG-GOP belonging to frame i, and k ∈ [0, n-1], l represent hierarchical location residing in the GOP of frame i belonging to it;
Step 1-2., for the referring-to relation across GOP, is defined as directly reference and one-level reference with reference to relation, indirect reference and secondary reference from low to high successively, three grades of references, level Four reference, Pyatyi reference ... reference level other definition by that analogy;
Step 1-3. defines weight factor α with what weigh direct reference affects intensity, 0< α <1, along with the increase of referring-to relation progression τ, the affecting intensity and will decay of reference, referring-to relation progression is for affecting intensity w (τ) definition as shown in formula (3) during τ:
w(τ)=α τ,τ=1,2,..,m (3)
Wherein, m is the highest with reference to relative degree number;
For needing RA coding structure to be processed, each GOP's in one of them BIG-GOP arrange by method described in step 2 to step 4;
Step 2. adds up referring-to relation number ∑ N (τ) of each GOP;
Be numbered all frames in the GOP of k and be designated as N by the sum of the τ level reference of the frame of other GOP k(τ); In the statistic processes of referring-to relation number, if a certain frame A in this GOP respectively by two different with reference to progression by a certain frame B reference of other GOP, then when adding up the referring-to relation number of this GOP, the minimum referring-to relation of a statistics frame A and frame B and do not record the referring-to relation of other ranks;
The GOP that step 3. is numbered k always affects intensity W kas shown in formula (4):
W k = &Sigma; &tau; = 1 m w k ( &tau; ) &CenterDot; N k ( &tau; ) = &Sigma; &tau; = 1 m &alpha; &tau; &CenterDot; N k ( &tau; ) - - - ( 4 )
Definition GOP affects strength retrogression's Graded factor Δ W ias shown in formula (5):
ΔW i=|W i-W i-1|,i=2,3,……,n-2,n-1 (5)
The QP-Offset of step 4.GOP level presents the rule increased progressively, and its concrete distribution method is as follows:
Step 4-1. is by Δ W inormalize to [0,1] interval and be designated as Δ W norm – i, as shown in formula (6):
ΔW norm-i=(ΔW i-ΔW 2)/(ΔW n-1-ΔW 2) (6)
Step 4-2. integrated encode quality and picture quality consider, and definition GOP level QP-Offset namely adjustable extent be [0,2];
The present invention is less than or equal to the situation of 2 to GOP number in BIG-GOP, does not make the amendment of the QP-Offset of GOP level;
If the number of GOP equals 3 in BIG-GOP, the adjustable extent of definition GOP level QP-Offset is [0,1]: directly the GOP level QP-Offset of definition the first two GOP is the GOP level QP-Offset of the zero, three GOP is 1;
If the number of GOP is greater than 4 in BIG-GOP, then perform step 4-3 to step 4-4;
If the number of GOP equals 4 in BIG-GOP, now GOP to affect intensity all less, be to ensure coding quality, side-play amount do not made to the first two GOP of BIG-GOP and changes, be i.e. the first two GOP in BIG-GOP all be set to zero, the side-play amount of residue GOP it is as follows that rule is set:
(1) if Δ W norm – i≤ 0.1, be then encoded to the GOP's of i be 1.
(2) if Δ W norm – i>0.1, be then numbered the GOP's of i be set to 2;
Step 4-3. sets self-defined minimum ε, and its span is (0,0.1);
If the number of GOP is greater than 4 in BIG-GOP, then the GOP level QP-Offset that the GOP being numbered zero in GOP and BIG-GOP at I-frame place distributes is set to zero; If be numbered the Δ W of 3 i.e. the 4th GOP norm – i< ε, being then numbered 1 and 2 is second and third GOP's be set to zero, otherwise be set to 1; The side-play amount that residue GOP is corresponding arrange by the rule that arranges described in this step 4-4;
Arranging of the GOP level QP-Offset of GOP is remained regular as follows in step 4-4.BIG-GOP:
(1) if Δ W norm – i< ε, be then numbered the GOP's of i be set to 0;
(2) if ε≤Δ W norm – i≤ 0.1, be then numbered the GOP's of i be set to 1;
(3) if Δ W norm – i>0.1, be then numbered the GOP's of i be set to 2;
Step 4-5., for needing RA coding structure to be processed, can obtain as stated above to each GOP's in each BIG-GOP the QP-Offset realizing the GOP level of RA coding structure is thus arranged.
2. the QP-Offset method to set up of GOP level according to claim 1, is characterized in that, the value of described weight factor α is 1/2.
CN201510341367.4A 2015-06-18 2015-06-18 A kind of QP Offset methods to set up of GOP levels Active CN104954793B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201510341367.4A CN104954793B (en) 2015-06-18 2015-06-18 A kind of QP Offset methods to set up of GOP levels

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201510341367.4A CN104954793B (en) 2015-06-18 2015-06-18 A kind of QP Offset methods to set up of GOP levels

Publications (2)

Publication Number Publication Date
CN104954793A true CN104954793A (en) 2015-09-30
CN104954793B CN104954793B (en) 2018-01-12

Family

ID=54169089

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201510341367.4A Active CN104954793B (en) 2015-06-18 2015-06-18 A kind of QP Offset methods to set up of GOP levels

Country Status (1)

Country Link
CN (1) CN104954793B (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2019100712A1 (en) * 2017-11-21 2019-05-31 腾讯科技(深圳)有限公司 Video encoding method, computer device and storage medium
CN111327896A (en) * 2018-12-13 2020-06-23 浙江宇视科技有限公司 Video transmission method and device, electronic equipment and readable storage medium
CN111901597A (en) * 2020-08-05 2020-11-06 杭州当虹科技股份有限公司 CU (CU) level QP (quantization parameter) allocation algorithm based on video complexity
CN112165618A (en) * 2020-09-27 2021-01-01 北京金山云网络技术有限公司 Video coding method, device, equipment and computer readable storage medium
CN112165620A (en) * 2020-09-24 2021-01-01 北京金山云网络技术有限公司 Video encoding method and device, storage medium and electronic equipment

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101572806A (en) * 2009-06-01 2009-11-04 北京邮电大学 Frame I code rate control method based on H264
CN101668197A (en) * 2009-09-18 2010-03-10 浙江大学 Code rate control method in scalable video coding based on linear model
US20110310963A1 (en) * 2009-12-14 2011-12-22 Hua Yang Method and apparatus for bit allocation in offline video coding

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101572806A (en) * 2009-06-01 2009-11-04 北京邮电大学 Frame I code rate control method based on H264
CN101668197A (en) * 2009-09-18 2010-03-10 浙江大学 Code rate control method in scalable video coding based on linear model
US20110310963A1 (en) * 2009-12-14 2011-12-22 Hua Yang Method and apparatus for bit allocation in offline video coding

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2019100712A1 (en) * 2017-11-21 2019-05-31 腾讯科技(深圳)有限公司 Video encoding method, computer device and storage medium
KR20200002032A (en) * 2017-11-21 2020-01-07 텐센트 테크놀로지(센젠) 컴퍼니 리미티드 Video encoding methods, computer devices and storage media
US10944970B2 (en) 2017-11-21 2021-03-09 Tencent Technology (Shenzhen) Company Limited Video coding method, computer device, and storage medium
KR102263979B1 (en) * 2017-11-21 2021-06-10 텐센트 테크놀로지(센젠) 컴퍼니 리미티드 Video encoding methods, computer devices and storage media
CN111327896A (en) * 2018-12-13 2020-06-23 浙江宇视科技有限公司 Video transmission method and device, electronic equipment and readable storage medium
CN111327896B (en) * 2018-12-13 2022-08-23 浙江宇视科技有限公司 Video transmission method and device, electronic equipment and readable storage medium
CN111901597A (en) * 2020-08-05 2020-11-06 杭州当虹科技股份有限公司 CU (CU) level QP (quantization parameter) allocation algorithm based on video complexity
CN112165620A (en) * 2020-09-24 2021-01-01 北京金山云网络技术有限公司 Video encoding method and device, storage medium and electronic equipment
CN112165618A (en) * 2020-09-27 2021-01-01 北京金山云网络技术有限公司 Video coding method, device, equipment and computer readable storage medium
CN112165618B (en) * 2020-09-27 2022-08-16 北京金山云网络技术有限公司 Video coding method, device, equipment and computer readable storage medium

Also Published As

Publication number Publication date
CN104954793B (en) 2018-01-12

Similar Documents

Publication Publication Date Title
CN104954793A (en) GOP-level QP-Offset setting method
CN100417229C (en) Coding apparatus, coding method, coding method program, and recording medium recording the coding method program
CN101711481B (en) Method and apparatus for video coding using prediction data refinement
CN104581162B (en) By using the method for large-scale converter unit decoded picture
CN104980744B (en) The method that video is decoded
RU2581553C2 (en) Method for encoding video using bias control according to pixel classification and device therefor, video decoding method and device therefor
CN101742321A (en) Layer decomposition-based Method and device for encoding and decoding video
CN101895756B (en) Method and system for coding, decoding and reconstructing video image blocks
CN1757240A (en) Video encoding
US20120014431A1 (en) Methods and Systems for Parallel Video Encoding and Parallel Video Decoding
CN102577379A (en) Methods and apparatus for embedded quantization parameter adjustment in video encoding and decoding
CN100574447C (en) Fast intraframe predicting mode selecting method based on the AVS video coding
US20120201294A1 (en) Methods for parallel video encoding and decoding
CN112075082A (en) Method and apparatus for video encoding and decoding for CABAC-based neural network implementation
CN109889852B (en) HEVC intra-frame coding optimization method based on adjacent values
CN113766247B (en) Loop filtering method and device
CN108259898B (en) Intra-frame fast coding method based on quality scalable video coding QSHVC
CN103260018A (en) Intra-frame image predictive encoding and decoding method and video codec
CN104159117A (en) Method for hiding HEVC video information
CN104796703B (en) The bit rate control method of scalable video based on predictive mode rate smoothing
CN105306957A (en) Adaptive loop filtering method and device
CN103634601A (en) Structural similarity-based efficient video code perceiving code rate control optimizing method
KR100785855B1 (en) Apparatus and method of embedded quantizaton for the improved SNR scalbilty
CN110139101B (en) Frame-level bit distribution method based on lambda domain code rate control
CN110677644B (en) Video coding and decoding method and video coding intra-frame predictor

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