CN101194510A - H.264量化 - Google Patents
H.264量化 Download PDFInfo
- Publication number
- CN101194510A CN101194510A CNA200680010777XA CN200680010777A CN101194510A CN 101194510 A CN101194510 A CN 101194510A CN A200680010777X A CNA200680010777X A CN A200680010777XA CN 200680010777 A CN200680010777 A CN 200680010777A CN 101194510 A CN101194510 A CN 101194510A
- Authority
- CN
- China
- Prior art keywords
- quantization
- matrix
- qmat
- integer
- coefficient
- 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.)
- Pending
Links
Images
Abstract
低复杂性(16位算法)H.264视频压缩用多个量化表取代用于所有量化参数的单个量化表,并借此使量化偏移和舍入加法相等;此消除了对32位存取的需要。
Description
技术领域
本发明涉及数字图像和视频信号处理,且更明确地说,涉及块变换和/或量化以及反量化和/或逆变换。
背景技术
存在用于数字视频通信和存储的各种应用,且已开发出并正在持续开发相应的国际标准。例如视频电话和会议的低位速率通信以及例如动画的较大视频文件压缩导致各种视频压缩标准:H.261、H.263、MPEG-1、MPEG-2、AVS等。这些压缩方法依赖于离散余弦变换(DCT)或类似变换以及变换系数的量化来减少需用来编码的位的数目。
基于DCT的压缩方法将图片分解为宏块,其中每一宏块含有四个8×8亮度块以及两个8×8色度块,但可使用其它块大小和变换变量。图2a描绘基于DCT的视频编码的功能块。为减小位速率,使用8×8 DCT将8×8的块(亮度和色度)转换为频域。接着,量化DCT系数的8×8的块,将其扫描到1-D序列中,并通过使用可变长度编码(VLC)进行编码。对于涉及运动补偿(MC)的预测性编码,需要反量化和IDCT用于反馈回路。除了MC之外,图2a中的所有功能块均基于8×8的块而操作。图2a中的速率控制单元负责在允许的范围内并根据目标位速率和缓冲器充满程度而产生量化步长(qp)来控制DCT系数量化单元。实际上,较大的量化步长暗示较多消没和/或较小的量化系数,这意味着较少和/或较短的代码字以及随之的较小的位速率和文件。
存在两种编码的宏块。内编码(INTRA-coded)的宏块独立于先前的参考帧而被编码。在中间编码(INTER-coded)的宏块中,首先针对(当前宏块的)每一块产生来自先前参考帧的经运动补偿的预测块,接着编码预测误差块(即,当前块与预测块之间的差异块)。
对于内编码的宏块来说,内编码的8×8DCT块中的第一(0,0)系数被称为DC系数,块中其余63个DCT系数是AC系数;而对于中间编码的宏块来说,中间编码的8×8DCT块的所有64个DCT系数均视为AC系数。可用固定值的量化步长来量化DC系数,而AC系数根据位速率控制来调节量化步长,所述位速率控制将迄今在编码图片过程中所使用的位与所分配的待使用的位的数目进行比较。此外,量化矩阵(例如,如在MPEG-4中)允许DCT系数之间变化的量化步长。
明确地说,8×8二维DCT定义为:
其中f(x,y)是输入的8×8样本块,且F(u,v)是输出的8×8变换块,其中u、v、x、y=0、1、...,7;且
应注意,此变换具有8×8矩阵乘法F=Dt×f×D的形式,其中“×”表示8×8矩阵的乘法且D是8×8矩阵,其中u、x元素等于
所述变换以双精度执行,且最终的变换系数舍入为整数值。接下来,将变换系数的量化定义为
其中QP是以双精度从量化步长qp中计算出的量化因子,作为一指数,例如:QP=2qp/6。量化的系数舍入为整数值且被编码。
相应的反量化变为:
F′(u,v)=QF(u,v)*QP
其中双精度值舍入为整数值。
最后,逆变换(重构样本块)为:
同样,双精度值舍入为整数值。
各种较新近的视频压缩方法(例如,H.264和AVS标准)通过使用整数变换取代DCT和/或不同大小的块而简化了双精度DCT方法。实际上,以类似于8×8 DCT变换系数矩阵D的元素来定义n×n整数变换矩阵Tn×n。接着,在fn×n和Fn×n分别表示输入的n×n样本数据矩阵(像素或余数的块)和输出的n×n变换系数块的情况下,将n×n整数正变换定义为:
Fn×n=Tt n×n×fn×n×Tn×n
其中“×”表示n×n矩阵乘法,且n×n矩阵Tt n×n是n×n矩阵Tn×n的转置矩阵。
举例来说,与其它现有视频标准中一样,在H.264中,最小的编码单位是宏块,其含有四个8×8亮度块以及来自两个色度分量的两个8×8色度块。然而,如图3所示,在H.264中,8×8块进一步被划分为4×4块以用于变换和量化,这导致宏块总共有二十四个4×4块。整数变换之后,来自两个色度分量中每一者的四个DC值合在一起以形成两个色度DC块,对所述两个色度DC块执行额外的2×2变换和量化。类似地,如果以INTRA 16×16模式编码宏块,那么十六个4×4亮度块的十六个DC值合在一起以创建4×4亮度DC块,对所述4×4亮度DC块实行4×4亮度DC变换和量化。
因此,在H.264中,存在三种变换和量化,即:针对二十四个亮度/色度块的4×4变换和量化;针对两个色度DC块的2×2变换和量化;以及当宏块被编码为INTRA 16×16模式时针对亮度DC块的4×4变换和量化。
经变换的系数的量化可如上所述为量化步长的指数,或可使用具有整数条目的查找表。反量化镜射量化。且逆变换也使用Tn×n,且其转置矩阵类似于使用D的DCT,且其转置矩阵用于正变换和逆变换两者。
因此,这些替代方法在维持性能的同时仍具有可减小的计算上的复杂性。
发明内容
本发明通过根据每个量化参数修改量化表而提供H.264图像/视频处理的低复杂性量化。
优选实施例方法提供有用于H.264视频编码中的简化的16位运算。
附图说明
图1a-1b是流程图。
图2a-2b说明具有DCT和其它变换及量化的运动补偿视频压缩。
图3展示H.264宏块结构。
图4说明方法的比较。
具体实施方式
优选实施例方法提供适用于16位H.264方法的简化的4×4和2×2变换的块量化。使量化查找表依赖于量化参数以使舍入与偏移相等;这避免了32位存取。
所述方法适用于用H.264整数变换以及变换系数的量化对(运动补偿的)像素块进行操作的视频压缩,其中量化可广泛变化。对于如图2b中所说明的H.264编码,来自位流输出缓冲器的缓冲器充满程度反馈可确定量化因子,其通常在1到200-500的范围内变化。优选实施例方法将适用于图2b中的块“量化”。图1a-1b是编码和解码的变换/量化流程。
优选实施例系统利用芯片上(SoC)数字信号处理器(DSP)或通用可编程处理器或特殊应用电路或系统(例如,具有RISC处理器控制的同一芯片上的DSP与RISC处理器两者)来执行优选实施例方法。具体来说,具有视频剪辑能力的数码相机(DSC)或具有视频能力的手机可包含优选实施例方法。所存储的程序可处于板上ROM或外部快闪EEPROM中,以供DSP或可编程处理器执行优选实施例方法的数字处理。模拟到数字转换器和数字到模拟转换器提供与真实世界的耦合,且调制器和解调器(以及用于无线接口的天线)为传输波形提供耦合。
首先,考虑到针对三种块类型的每一者的H.264变换、量化及其相反过程:4×4亮度/色度块、2×2色度DC块和4×4亮度DC块;优选实施例方法提供H.264的量化的简化。
(a)针对4×4亮度/色度块的正变换
4×4正变换使用以下4×4变换矩阵T4×4用于与宏块的二十四个4×4亮度/色度块中的每一4×4样本数据矩阵进行矩阵乘法:
因此,具有元素xij的4×4矩阵到具有元素yij的4×4矩阵的正变换为
应注意,T4×4的列正交,且T4×4大致上与4×4 DCT矩阵成比例。
(b)针对4×4亮度/色度块的量化
yij(其中,i=0、1、2、3且j=0、1、2、3)经量化以给出cij作为量化参数qP的函数:
cij=sign(yij)*(|yij|*QLevelScale(qP%6,i,j)+Δ)>>(15+qP/6)
其中QLevelScale(qP%6,i,j)是量化查找表;qP表示亮度量化参数QPY或色度量化参数QPC(QPY和QPC均在0、1、...,53的范围内);Δ=α*215+qP/6,其中舍入参数0<α<1;符号(.)是符号函数(z为正时sign(z)=+1,z为负时sign(z)=-1,且sign(0)=0);*表示标量乘法;/是整数除法(整数商且余数舍弃);%是模运算,其实质上是从整数除法舍弃的余数;且》和《表示右移和左移,其适用于以二进制记数法表示的数字。应注意,qP/6处于0到8的范围内。量化查找表由六个4×4的比例矩阵组成,每一者用于qP%6的六个可能值中的每一值。每一4×4比例矩阵具有相同的简单形式但不同的元素值:
QLevelScale[6][4][4]=
{
{{13107,8066,13107,8066 },{8066,5243,8066,5243},{13107,8066,13107,8066},{8066,5243,8066,5243}},
{{11916,7490,11916,7490},{7490,4660,7490,4660},{11916,7490,11916,7490},{7490,4660,7490,4660}},
{{10082,6554,10082,6554},{6554,4194,6554,4194},{10082,6554,10082,6554},{6554,4194,6554,4194}},
{{9362,5825,9362,5825},{5825,3647,5825,3647},{9362,5825,9362,5825},{5825,3647,5825,3647}},
{{8192,5243,8192,5243},{5243,3355,5243,3355},{8192,5243,8192,5243},{5243,3355,5243,3355}},
{{7282,4559,7282,4559},{4559,2893,4559,2893},{7282,4559,7282,4559},{4559,2893,4559,2893}}
};
应注意,总体上,所述量化大致为:通过与211与214之间的整数比例因子进行乘法,随后与215进行整数除法(其补偿整数比例因子的大小),且接着是与处于1到28范围内并为量化提供位数目的减小的2qP/6进行整数除法。经量化的系数cij最终被编码并传输/存储。
(c)针对4×4亮度/色度块的反量化
在解码以恢复cij之后,针对4×4量化块cij(其中,i=0、1、2、3且j=0、1、2、3)的反量化给出dij为:
dij=(cij*IQLevelScale(qP%6,i,j))<<qP/6
其中,同样地,qP表示亮度量化参数QPY或色度量化参数QPC,且IQLevelScale(qP%6,i,j)是反量化查找表条目。反量化查找表同样由针对六个可能的qP%6中的每一者的4×4比例矩阵组成,其中每一4×4比例矩阵具有四个低值元素,八个中间值元素和四个高值元素:
IQLevelScale[6][4][4]=
{
{{10,13,10,13},{13,16,13,16},{10,13,10,13},{13,16,13,16}},
{{11,14,11,14},{14,18,14,18},{11,14,11,14},{14,18,14,18}},
{{13,16,13,16},{16,20,16,20},{13,16,13,16},{16,20,16,20}},
{{14,18,14,18},{18,23,18,23},{14,18,14,18},{18,23,18,23}},
{{16,20,16,20},{20,25,20,25},{16,20,16,20},{20,25,20,25}},
{{18,23,18,23},{23,29,23,29},{18,23,18,23},{23,29,23,29}}
};
应注意,左移提供量化期间与2qP/6进行整数除法中损失的数目的位的恢复,且通过在量化过程中与QLevelScale(qP%6,i,j)进行乘法以及除以215而导致的先前的量值减小实质上抵消与IQLevelScale(qP%6,i,j)进行乘法而导致的量值增加。
(d)针对4×4亮度/色度块的逆变换
逆4×4变换与DCT不同之处在于,4×4变换矩阵的转置矩阵不等于4×4矩阵的逆矩阵,因为行具有不同范数;也就是说,T4×4不是正交矩阵。实际上,量化和反量化的比例矩阵调节经变换像素的相对大小。明确地说,逆变换使用4×4矩阵V4×4及其转置矩阵,其中:
应注意,V4×4看上去和Tt 4×4一样,但两列按比例缩放了1/2以减小动态范围。因此,具有元素dij(其中i=0、1、2、3且j=0、1、2、3)的4×4矩阵的逆变换是具有元素hij的4×4矩阵,其定义为:
最后,hij按比例缩减为rij=(hij+32)>>6以定义恢复的(经解码和解压缩的)数据。
类似的变换和量化适用于2×2色度DC块。
(e)针对2×2色度DC块的正变换
正2×2变换使用以下2×2变换矩阵T2×2与宏块的两个2×2色度DC块的每一2×2样本数据矩阵进行矩阵乘法:
因此,具有元素xij的2×2矩阵到具有元素yij的2×2矩阵的正变换为
(f)针对2×2色度DC块的量化
yij(其中,i=0、1且j=0、1)经量化以给出cij作为量化参数QPc的函数:
cij=sign(yij)*(|yij|*QLevelScale(QPC%6,0,0)+Δ)>>(16+QPC/6)
其中QLevelScale(QPC%6,0,0)是上文(b)中列出的量化查找表中的条目;QPC与之前一样是色度量化因子且在0、1、...,51的范围内;且Δ={α*216+QPc/6,其中舍入参数0<α<1。这些经量化的系数cij最终被编码并传输/存储。
(g)针对2×2色度DC块的逆变换
在解码以恢复2×2经量化的DC块cij(其中,i=0、1且j=0、1)之后,在反量化之前进行逆2×2变换以给出fij为:
应注意,与DCT一样,所述变换实质上是其自身的逆矩阵。
(h)针对2×2色度DC块的反量化
fij(其中,i=0、1且j=0、1)经反量化以给出dcCij作为量化参数QPc的函数:
dcCij=((fij*IQLevelScale(QPC%6,0,0))<<QPC/6)>>1
其中,同样,QPC表示色度量化参数,且IQLevelScale(qP%6,0,0)是(c)中列出的反量化查找表的(0,0)条目。
最后,类似的变换和量化适用于4×4亮度DC块。
(i)针对4×4亮度DC块的正变换
4×4亮度DC块xij到具有元素hij的4×4矩阵的正变换为
接着按比例缩放hij以使yij=(hij+1)>>1而得到变换yij。
(j)针对4×4亮度DC块的量化
yij(其中,i=0、1、2、3且j=0、1、2、3)经量化以给出cij作为亮度量化参数QPY的函数:
cij=sign(yij)*(|yij|*QLevelScale(QPY%6,0,0)+Δ)>>(16+QPY/6)
其中QLevelScale(QPy%6,0,0)是(b)中列出的量化查找表中的(0,0)条目;且同样Δ=α*216+Qpy/6,其中0<α<1是舍入参数。
(k)针对4×4亮度DC块的逆变换
在解码以恢复4×4经量化DC块cij(其中,i=0、1、2、3且j=0、1、2、3)之后,在反量化之前进行逆4×4变换以给出fii为:
(l)针对4×4亮度DC块的反量化
fij(其中,i=0、1、2、3且j=0、1、2、3)经反量化以给出dcYij作为量化参数QPY的函数:
dcYij=((fij*IQLevelScale(QPY%6,0,0))<<(QPY/6)+2)>>2
其中,同样地,QPY表示亮度量化参数,且IQLevelScale(QPY%6,0,0)是(c)中列出的反量化查找表的(0,0)条目。
在开发H.264标准期间,努力确保H.264变换和量化可以16位算法实施。这一目标已基本上实现。然而,以上步骤(b)、(f)和(j)的正量化中所使用的舍入控制参数Δ可能超过16位;且这使得所实施的H.264正量化在不具有32位存储器存取的处理器上不能实行。实际上,Δ=α*215+qP/6或α*216+qP/6,其可高达24位。因此,优选实施例提供具有恒定Δ的H.264正量化。明确地说,针对4×4亮度/色度块的变换以及量化及其相反过程的优选实施例方法使用以上步骤(a)、(c)和(d),但用新的步骤(b′)代替步骤(b);对于2×2色度DC块,使用以上步骤(e)、(g)和(h),但用新的步骤(f′)代替步骤(f);且对于4×4亮度DC块,使用以上步骤(i)、(k)和(l),但用新的步骤(j′)代替步骤(j)。这些新的步骤如下:
(b′)针对4×4亮度/色度块的优选实施例量化
yij(其中,i=0、1、2、3且j=0、1、2、3)经量化以给出cij作为量化参数qP的函数:
cij=sign(yij)*(|yij|*QMat(0)(qP%6,i,j)+α*215)>>15(qP/6=0时)
cij=sign(yij)*(|yij|*QMat(qP%6-1)(qP%6,i,j)+α*216)>>16 (qP/6≠0时)
其中,与(b)中一样,qP表示亮度量化参数QPY或色度量化参数QPc′,且也与(b)中一样,0<α<1是舍入参数。QMat(n)(qP%6,i,j)是依据(b)中列出的QLevelScale(qP%6,i,j)定义的新的量化查找表,且定义为:
QMat(0)(qP%6,i,j)=QLevelScale(qP%6,i,j)
QMat(n)(qP%6,i,j)=(QLevelScale(qP%6,i,j)+2n-1)>>n (其中,n>0)
也就是说,依据qP/6,QLevelScale[6][4][4]被QMat(0)[6][4][4]、QMat(1)[6][4][4]、...,或QMat(7)[6][4][4]取代。应注意,对于QMat(n)[6][4][4]条目,相应的QLevelScale[6][4][4]条目存在舍入n位的右移;右移将条目的大小从范围211-214减小到范围211-n-214-n。(应注意,0到8范围内的qP/6暗示着n将在范围1到7内。)这样使用较多的表允许大小依赖于qP/6的Δ被恒定大小的α*216(或者当qP/6=0时,为α*215)取代,其为16位整数。举例来说,表QLevelScale(0,i,j)的三个不同的值是13107、8066和5243;而QMat(7)(0,i,j)的相应条目分别为102、63和41。这通过与较低的解析度折衷而节省了7位。
(f′)针对2×2色度DC块的优选实施例量化
yij(其中,i=0、1且j=0、1)经量化以给出cij作为色度量化参数QPc的函数:
cij=sign(yij)*(|yij|*QMat(Qpc/6)(QPc%6,0,0)+α*216)>>16
其中(b′)定义了QMat(QPc/6)(QPC%6,0,0)和α。应注意,还需要QMat(8)(QPC%6,0,0);而(b′)仅使用QMat(n)(qP%6,i,j)(其中n≤7)。
(j′)针对4×4亮度DC块的优选实施例量化
yij(其中,i=0、1、2、3且j=0、1、2、3)经量化以给出cij作为亮度量化参数QPY的函数:
cij=sign(yij)*(|yij|*QMat(Qpy/6)(QPY%6,0,0)+α*216)>>16
其中(b′)定义了QMat(QPy/6)(QPY%6,0,0)和α。同样应注意,还需要QMat(8)(QPY%6,0,0)。
对于表大小并非考虑因素的实施方案,可预先计算并存储(b′)、(f′)和(j′)中所使用的新的量化矩阵。确切的新的量化矩阵如下:
□QMat(0)[6][4][4]={
{{13107,8066,13107,8066},{8066,5243,8066,5243},{13107,8066,13107,8066},{8066,5243,8066,5243}},
{{11916,7490,11916,7490},{7490,4660,7490,4660},{11916,7490,11916,7490},{ 7490,4660,7490,4660}},
{{10082,6554,10082,6554},{6554,4194,6554,4194},{10082,6554,10082,6554},{6554,4194,6554,4194}},
{{9362,5825,9362,5825},{5825,3647,5825,3647},{9362,5825,9362,5825},{5825,3647,5825,3647}},
{{8192,5243,8192,5243},{5243,3355,5243,3355},{8192,5243,8192,5243},{5243,3355,5243,3355}},
{{7282,4559,7282,4559},{4559,2893,4559,2893},{7282,4559,7282,4559},{4559,2893,4559,2893}},
};
QMat(1)[6][4][4]={
{{6554,4033,6554,4033},{4033,2622,4033,2622},{6554,4033,6554,4033},{4033,2622,4033,2622}},
{{5958,3745,5958,3745},{3745,2330,3745,2330},{5958,3745,5958,3745},{3745,2330,3745,2330}},
{{5041,3277,5041,3277},{3277,2097,3277,2097},{5041,3277,5041,3277},{3277,2097,3277,2097}},
{{4681,2913,4681,2913},{2913,1824,2913,1824},{4681,2913,4681,2913},{2913,1824,2913,1824}},
{{4096,2622,4096,2622},{2622,1678,2622,1678},{4096,2622,4096,2622},{2622,1678,2622,1678}},
{{3641,2280,3641,2280},{2280,1447,2280,1447},{3641,2280,3641,2280},{2280,1447,2280,1447}},
};
QMat(2)[6][4][4]={
{{3277,2017,3277,2017},{2017,1311,2017,1311},{3277,2017,3277,2017},{2017,1311,2017,1311}},
{{2979,1873,2979,1873},{1873,1165,1873,1165},{2979,1873,2979,1873},{1873,1165,1 873,1165}},
{{2521,1639,2521,1639},{1639,1049,1639,1049},{2521,1639,2521,1639},{1639,1049,1639,1049}},
{{2341,1456,2341,1456},{1456,912,1456,912},{2341,1456,2341,1456},{1456,912,1456,912}},
{{2048,1311,2048,1311},{1311,839,1311,839},{2048,1311,2048,1311},{1311,839,1311,839}},
{{1821,1140,1821,1140},{1140,723,1140,723},{1821,1140,1821,1140},{1140,723,1140,723}},
};
QMat(3)[6][4][4]={
{{1638,1008,1638,1008},{1008,655,1008,655},{1638,1008,1638,1008},{1008,655,1008,655}},
{{1490,936,1490,936},{936,583,936,583},{1490,936,1490,936},{936,583,936,583}},
{{1260,819,1260,819},{819,524,819,524},{1260,819,1260,819},{819,524,819,524}},
{{1170,728,1170,728},{728,456,728,456},{1170,728,1170,728},{728,456,728,456}},
{{1024,655,1024,655},{655,419,655,419},{1024,655,1024,655},{655,419,655,419}},
{{910,570,910,570},{570,362,570,362},{910,570,910,570},{570,362,570,362}},
};
QMat(4)[6][4][4]={
{{819,504,819,504},{504,328,504,328},{819,504,819,504},{504,328,504,328}},
{{745,468,745,468},{468,291,468,291},{745,468,745,468},{468,291,468,291}},
{{630,410,630,410},{410,262,410,262},{630,410,630,410},{410,262,410,262}},
{{585,364,585,364},{364,228,364,228},{585,364,585,364},{364,228,364,228}},
{{512,328,512,328},{328,210,328,210},{512,328,512,328},{328,210,328,210}},
{{455,285,455,285},{285,181,285,181},{455,285,455,285},{285,181,285,181}},
};
QMat(5)[6][4][4]={
{{410,252,410,252},{252,164,252,164},{410,252,410,252},{252,164,252,164}},
{{372,234,372,234},{234,146,234,146},{372,234,372,234},{234,146,234,146}},
{{315,205,315,205},{205,131,205,131},{315,205,315,205},{205,131,205,131}},
{{293,182,293,182},{182,114,182,114},{293,182,293,182},{182,114,182,114}},
{{256,164,256,164},{164,105,164,105},{256,164,256,164},{164,105,164,105}},
{{228,142,228,142},{142,90,142,90},{228,142,228,142},{142,90,142,90}},
};
QMat(6)[6][4][4]={
{{205,126,205,126},{126,82,126,82},{205,126,205,126},{126,82,126,82}},
{{186,117,186,117},{117,73,117,73},{186,117,186,117},{117,73,117,73}},
{{158,102,158,102},{102,66,102,66},{158,102,158,102},{102,66,102,66}},
{{146,91,146,91},{91,57,91,57},{146,91,146,91},{91,57,91,57}},
{{128,82,128,82},{82,52,82,52},{128,82,128,82},{82,52,82,52}},
{{114,71,114,71},{71,45,71,45},{114,71,114,71},{71,45,71,45}},
};
QMat(7)[6][4][4]={
{{102,63,102,63},{63,41,63,41},{102,63,102,63},{63,41,63,41}},
{{93,59,93,59},{59,36,59,36},{93,59,93,59},{59,36,59,36}},
{{79,51,79,51},{51,33,51,33},{79,51,7 9,51},{51,33,51,33}},
{{73,46,73,46},{46,28,46,28},{73,46,73,46},{46,28,46,28}},
{{64,41,64,41},{41,26,41,26},{64,41,64,41},{41,26,41,26}},
{{57,36,57,36},{36,23,.36,23},{57,36,57,36},{36,23,36,23}},
};
QMat(8)[6][4][4]={
{{51,32,51,32},{32,20,32,20},{51,32,51,32},{32,20,32,20}},
{{4 7,29,47,29},{29,18,29,18},{47,29,47,29},{29,18,29,18}},
{{39,26,39,26},{26,16,26,16},{39,26,39,26},{26,16,26,16}},
{{37,23,37,23},{23,14,23,14},{37,23,37,23},{23,14,23,14}},
{{32,20,32,20},{20,13,20,13},{32,20,32,20),{20,13,20,13}},
{{28,18,28,18},{18,11,18,11},{28,18,28,18},{18,11,18,11}},
};
应注意,在QMat(8)[6][4][4]中,仅使用QMat(8)(0,0,0)、QMat(8)(1,0,0)、QMat(8)(2,0,0)、QMat(8)(3,0,0)、QMat(8)(4,0,0)、QMat(8)(5,0,0),QMat(8)[6][4][4]中的其余分量不需要存储。因此,总的表大小为约1350字节(QMat(0)到QMat(5)存储为二字节条目,QMat(6)到QMat(8)存储为一字节条目)。
对于需要较小表大小的实施方案,可根据量化比例QPY和QPC通过以下计算而在运行中计算宏块的量化矩阵
QMat(0)(QPy%6,i,j)=QLevelScale(QPY%6,i,j)(其中QPY/6<2)
QMat(QPy/6-1)(QPY%6,i,j)=
(QLevelScale(QPY%6,i,j)+2Qpy/6-2)>>(QPY/6-1)(其中QPY/6≮2)
QMat(0)(QPC%6,i,j)=QLevelScale(QPC%6,i,j)(其中QPC/6<2)
QMat(QPc/6-1)(QPC%6,i,j)=
(QLevelScale(QPC%6,i,j)+2QPc/6-2)>>(QPC/6-1)(其中QPC/6≮2)
且
QMat(QPy/6)(QPY%6,0,0)=(QLevelScale(QPY%6,0,0)+2QPy/6-1)>>QPY/6
QMat(QPc/6)(QPY%6,0,0)=(QLevelScale(QPC%6,0,0)+2QPc/6-1)>>QPC/6
因此,对于宏块,需要计算16个亮度块的4×4量化矩阵、8个色度块的4×4量化矩阵、4×4亮度DC块的量化比例,以及两个2×2色度DC块的量化比例,以用于根据给定的QPY和QPC进行变换系数量化。由于量化比例对于不同的宏块并不非常频繁地改变,所以通常不需要对每一宏块执行此类量化矩阵计算。
实行模拟以测试用于H264的优选实施例简化的正量化的效率。“Anchor T&Q”是H264变换以及量化,其由等式(a)到(l)组成,“Simplified T&Q”由等式(a)、(b′)、(c)、(d)、(e)、(f′)、(g)、(h)、(i)、(j′)、(k)和(l)组成;也就是说,在此情况下,仅正量化发生变化,其它所有部分均保持不变。测试所有的量化步长(qp=0、1、2、...51)。用5000个随机宏块测试每一qp,样本值在[-255:255]的范围内。针对每一qp的所有测试样本宏块计算(见图4)输入的样本宏块与其重构的宏块之间的PSNR值。下表1、2、3中列出结果。
qp | Anchor T&Q SNR0Y,U,V[dB] | Simplified T&Q SNR1Y,U,V[dB],(ΔdB) |
0 | 59.105,59.168,59.180 | 59.105,59.168,59.180(0.000,0.000,0.000) |
1 | 57.688,57.765,57.697 | 57.688,57.765,57.697(0.000,0.000,0.000) |
2 | 55.667,55.679,55.683 | 55.667,55.679,55.683(0.000,0.000,0.000) |
3 | 55.094,55.119,55.125 | 55.094,55.119,55.125 (0,000,0.000,0.000) |
4 | 53.950,53.986,53.999 | 53.950,53.986,53.999(0.000,0.000,0.000) |
5 | 52.975,52.997,52.981 | 52.975,52.997,52.981(0.000,0.000,0.000) |
6 | 51.955,52.020,51.998 | 51.955,52.020,51.998(0.000,0.000,0.000) |
7 | 51.344,51.363,51.412 | 51.344,51.363,51.412(0.000,0.000,0.000) |
8 | 50.505,50.545,50.539 | 50.505,50.545,50.539(0.000,0.000,0.000) |
9 | 49.860,49.921,49.904 | 49.860,49.921,49.904(0.000,0.000,0.000) |
10 | 48.933,49.035,48.984 | 48.933,49.035,48.984(0.000,0.000,0.000) |
11 | 47.861,47.912,47.900 | 47.861,47.912,47.900(0.000,0.000,0.000) |
12 | 46.876,46.978,46.887 | 46.876,46.978,46.887(0.000,0.000,0.000) |
13 | 45.942,46.011,46.128 | 45.942,46.011,46.128(0.000,0.000,0.000) |
14 | 45.202,45.192,45.258 | 45.202,45.192,45.258(0.000,0.000,0.000) |
15 | 44.111,44.224,44.244 | 44.127,44.224,44.245(0.016,0.000,0.000) |
16 | 43.187,43.371,43.340 | 43.203,43.378,43.346(0.016,0.007,0.007) |
17 | 42.303,42.242,42.273 | 42.322,42.242,42.274(0.019,0.000,0.000) |
18 | 41.107,41.163,41.189 | 41.162,41.164,41.190(0.054,0.001,0.001) |
19 | 40.248,40.372,40.340 | 40.295,40.372,40.340(0.046,0.000,0.000) |
20 | 39.175,39.257,39.275 | 39.210,39.262,39.281(0.035,0.005,0.005) |
21 | 38.179,38.323,38.365 | 38.186,38.323,38.365(0.007,0.000,0.000) |
22 | 37.418,37.420,37.515 | 37.428,37.420,37.515(0.010,0.000,0.000) |
23 | 36.424,36.381,36.335 | 36.414,36.368,36.323(-0.010,-0.013,-0.012) |
24 | 35.103,35.323,35.289 | 35.100,35.323,35.289(-0.003,0.000,0.000) |
25 | 34.361,34.594,34.562 | 34.376,34.596,34.563(0.015,0.001,0.001) |
26 | 33.240,33.474,33.466 | 33.239,33.474,33.466(-0.001,0.000,0.000) |
27 | 32.243,32.521,32.505 | 32.249,32.521,32.505(0.006,0.000,0.000) |
28 | 31.335,31.617,31.650 | 31.324,31.616,31.649(-0.011,-0.001,-0.001) |
29 | 30.276,30.548,30.556 | 30.254,30.539,30.531(-0.022,-0.008,-0.025) |
30 | 29.276,30.557,30.580 | 29.283,30.547,30.552(0.007,-0.009,-0.028) |
31 | 28.491,29.604,29.628 | 28.481,29.604,29.628(-0.010,0.000,0.000) |
32 | 27.300,28.922,28.890 | 27.321,28.922,28.889(0.021,-0.000,-0.000) |
33 | 26.371,27.819,27.840 | 26.376,27.819,27.840(0.005,0.000,0.000) |
34 | 25.459,27.798,27.856 | 25.470,27.798,27.856(0.012,0.000,0.000) |
35 | 24.216,26.944,26.933 | 24.208,26.945,26.933(-0.008,0.000,0.000) |
36 | 23.307,26.169,26.158 | 23.316,26.169,26.158(0.009,0.000,0.000) |
37 | 22.623,26.176,26.144 | 22.638,26.176,26.144(0.015,0.000,0.000) |
38 | 21.477,25.190,25.223 | 21.476,25.187,25.220(-0.001,-0.003,-0.003) |
39 | 20.498,25.180,25.218 | 20.501,25.177,25.215(0.003,-0.003,-0.003) |
40 | 19.688,24.510,24.473 | 19.695,24.510,24.473(0.007,0.000,0.000) |
41 | 18.517,24.505,24.488 | 18.488,24.505,24.488(-0.029,0 000,0.000) |
42 | 17.527,24.006,24.017 | 17.528,24.008,24.021(0.001,0.002,0.004) |
43 | 16.823,24.022,24.009 | 16.832,24.025,24.013(0.008,0.004,0.004) |
44 | 15.786,24.006,24.023 | 15.789,24.009,24.026(0.003,0.004,0.003) |
45 | 14.903,23.439,23.448 | 14.898,23.440,23.449(-0.004,0.001,0.000) |
46 | 14.141,23.447,23.445 | 14.123,23.447,23.446(-0.018,0.000,0.001) |
47 | 13.161,23.449,23.450 | 13.144,23.450,23.451(-0.018,0.001,0.001) |
48 | 12.462,23.124,23.119 | 12.459,23.124,23.119(-0.004,-0.000,-0.000) |
49 | 11.969,23.126,23.116 | 11.966,23.126,23.116(-0.003,-0.001,-0.000) |
50 | 11.404,23,125,23.120 | 11.411,23.124,23.120(0.006,-0.000,-0.000) |
51 | 11.086,23.129,23.126 | 11.088,23.129,23.126(0.002,-0.000,-0.000) |
表1.模拟结果,中间编码的宏块,使用α=1/6
qp | Anchor T&Q SNR0Y,U,V[dB] | Simplified T&Q SNR1Y,U,V[dB],(ΔdB) |
0 | 65.655,65.735,65.687 | 65.655,65.735,65.687(0.00 0,0.000,0.000) |
1 | 63.130,63.198,63.415 | 63.130,63.198,63.415(0.000,0.000,0.000) |
2 | 60.202,60.213,60.218 | 60.202,60.213,60.218(0.000,0.000,0.000) |
3 | 58.500,58.699,58.704 | 58.500,58.699,58.704(0.000,0.000,0.000) |
4 | 57.331,57.158,57.197 | 57.331,57.158,57.197(0.000,0.000,0.000) |
5 | 55.651,55.646,55.650 | 55.651,55.646,55.650(0.000,0.000,0.000) |
6 | 54.264,54.293,54.305 | 54.264,54.293,54.305(0.000,0.000,0.000) |
7 | 53.502,53.502,53.543 | 53.502,53.502,53.543(0.000,0.000,0.000) |
8 | 52.686,52.692,52.711 | 52.686,52.692,52.711(0.000,0.000,0.000) |
9 | 51.794,51.813,51.792 | 51.794,51.813,51.792(0.000,0.000,0.000) |
10 | 50.804,51.011,50.933 | 50.804,51.011,50.933(0.000,0.000,0.000) |
11 | 50.138,50.132,50.132 | 50.138,50.132,50.132(0.000,0.000,0.000) |
12 | 49.182,49.181,49.162 | 49.182,49.181,49.162(0.000,0.000,0.000) |
13 | 48.321,48.407,48.327 | 48.321,48.407,48.327(0.000,0.000,0.000) |
14 | 47.325,47.331,47.320 | 47.325,47.331,47.320(0.000,0.000,0.000) |
15 | 46.247,46.358,46.371 | 46.393,46.496,46.512(0.146,0.138,0.141) |
16 | 45.603,45.594,45.608 | 45.617,45.594,45.608(0.014,0.000,0.000) |
17 | 44.544,44.469,44.494 | 44.564,44.473,44.497(0.020,0.004,0.003) |
18 | 43.394,43.418,43.465 | 43.409,43.418,43.465(0.015,0.000,0.000) |
19 | 42.715,42.716,42.677 | 42.717,42.716,42.677(0.002,0.000,0.000) |
20 | 41.581,41.579,41.589 | 41.596,41.579,41.590(0.016,0.000,0.001) |
21 | 40.548,40.596,40.569 | 40.554,40.596,40.569(0.006,0.000,0.000) |
22 | 39.686,39.598,39.735 | 39.693,39.598,39.735(0.008,0.000,0.000) |
23 | 38.765,38.547,38.585 | 38.756,38.534,38.574(-0.009,-0.013,-0.011) |
24 | 37.544,37.576,37.570 | 37.530,37.575,37.570(-0.014,-0.000,0.000) |
25 | 36.778,36.818,36.770 | 36.802,36.836,36.787(0.024,0.01 8,0.017) |
26 | 35.637,35.607,35.639 | 35.634,35.607,35.639(-0.002,0.000,0.000) |
27 | 34.527,34.667,34.617 | 34.532,34.667,34.61 8(0.004,0.000,0.000) |
28 | 33.702,33.787,33.768 | 33.678,33.785,33.767(-0.025,-0.001,-0.001) |
29 | 32.453,32.616,32.626 | 32.436,32.591,32.597(-0.017,-0.025,-0.029) |
30 | 31.571,32.609,32.638 | 31.581,32.586,32.610(0.010,-0.023,-0.028) |
31 | 30.831,31.631,31.638 | 30.829,31.631,31.638(-0.002,0.000,0.000) |
32 | 29.640,30.845,30.868 | 29.640,30.845,30.868(-0.001,0.000,0.000) |
33 | 28.587,29.740,29.699 | 28.589,29.740,29.699(0.002,0.000,0.000) |
34 | 27.747,29.733,29.703 | 27.763,29.733,29.703(0.016,0.000,-0.000) |
35 | 26.558,28.759,28.740 | 26.549,28.759,28.740(-0.010,0.000,0.000) |
36 | 25.529,27.850,27.867 | 25.531,27.850,27.867(0.002,0.000,0.000) |
37 | 24.762,27.857,27.857 | 24.775,27.857,27.857(0.013,0.000,0.000) |
38 | 23.652,26.742,26.743 | 23.660,26.729,26.734(0.008,-0.013,-0.009) |
39 | 22.650,26.737,26.736 | 22.653,26.723,26.726(0.003,-0.013,-0.010) |
40 | 21.785,25.859,25.851 | 21.788,25.859,25.851(0.003,0.000,0.000) |
41 | 20.583,25.859,25.853 | 20.552,25.859,25.853(-0.031,0.000,0.000) |
42 | 19.626,25.203,25.207 | 19.631,25.207,25.212(0.005,0.004,0.005) |
43 | 18.838,25.216,25.211 | 18.841,25.220,25.215(0.003,0.004,0.004) |
44 | 17.655,25.196,25.221 | 17.659,25.200,25.226(0.004,0.004,0.005) |
45 | 16.697,24.358,24.371 | 16.694,24.359,24.372(-0.004,0.001,0.001) |
46 | 15.818,24.362,24.349 | 15.809,24.363,24.350(-0.009,0.001,0.001) |
47 | 14.711,24.368,24.343 | 14.692,24.369,24.344(-0.019,0.001,0.001) |
48 | 13.844,23.798,23.796 | 13.842,23.798,23.795(-0.003,-0.000,-0.001) |
49 | 13.186,23.801,23.785 | 13.195,23.801,23.785(0.009,-0.001,-0.001) |
50 | 12.313,23.789,23.790 | 12.315,23.788,23.789(0.002,-0.001,-0.001) |
51 | 11.751,23.807,23.792 | 11.755,23.806,23.791(0.003,-0.001,-0.000) |
表2.模拟结果,INTRA4×4编码的宏块,使用α=1/3
qp | Anchor T&Q SNR0Y,U,V[dB] | Simplified T&Q SNR1Y,U,V[dB],(ΔdB) |
0 | 65.578,65.735,65.687 | 65.578,65.735,65.687(0.000,0.000,0.000) |
1 | 63.099,63.198,63.415 | 63.099,63.198,63.415(0.000,0.000,0.000) |
2 | 60.191,60.213,60.218 | 60.191,60.213,60.218(0.000,0.000,0.000) |
3 | 58.494,58.699,58.704 | 58.494,58.699,58.704(0.000,0.000,0.000) |
4 | 57.281,57.158,57.197 | 57.281,57.158,57.197(0.000,0.000,0.000) |
5 | 55.644,55.646,55.650 | 55.644,55.646,55.650(0.000,0.000,0.000) |
6 | 54.299,54.293,54.305 | 54.299,54.293,54.305(0.000,0.000,0.000) |
7 | 53.497,53.502,53.543 | 53.497,53.502,53.543(0.000,0.000,0.000) |
8 | 52.686,52.692,52.711 | 52.686,52.692,52.711(0.000,0.000,0.000) |
9 | 51.770,51.813,51.792 | 51.770,51.813,51.792(0.000,0.000,0.000) |
10 | 50.867,51.011,50.933 | 50.867,51.011,50.933(0.000,0.000,0.000) |
11 | 50.123,50.132,50.132 | 50.123,50.132,50.132(0.000,0.000,0.000) |
12 | 49.153,49.181,49.162 | 49.153,49.181,49.162(0.000,0.000,0.000) |
13 | 48.334,48.407,48.327 | 48.334,48.407,48.327(0.000,0.000,0.000) |
14 | 47.312,47.331,47.320 | 47.312,47.331,47.320(-0.000,0.000,0.000) |
15 | 46.255,46.358,46.371 | 46.403,46.496,46.512(0.148,0.138,0.141) |
16 | 45.545,45.594,45.608 | 45.559,45.594,45.608(0.014,0.000,0.000) |
17 | 44.530,44.469,44.494 | 44.550,44.473,44.497(0.020,0.004,0.003) |
18 | 43,401,43.418,43.465 | 43.416,43.418,43.465(0.015,0.000,0.000) |
19 | 42.705,42.716,42.677 | 42.709,42.716,42.677(0.003,0.000,0.000) |
20 | 41.599,41.579,41.589 | 41.615,41.579,41.590(0.015,0.000,0.001) |
21 | 40.544,40.596,40.569 | 40.550,40.596,40.569(0.006,0.000,0.000) |
22 | 39.732,39.598,39.735 | 39.740,39.598,39.735(0.008,0.000,0.000) |
23 | 38.757,38.547,38.585 | 38.737,38.534,38.574(-0.021,-0.013,-0.011) |
24 | 37.528,37.576,37.570 | 37.515,37.575,37.570(-0.014,-0.000,0.000) |
25 | 36.785,36.818,36.770 | 36.810,36.836,36.787(0.024,0.018,0.017) |
26 | 35.642,35.607,35.639 | 35.640,35.607,35.639(-0.002,0.000,0.000) |
27 | 34.536,34.667,34.617 | 34.540,34.667,34.61 8(0.004,0.000,0.000) |
28 | 33.725,33.787,33.768 | 33.700,33.785,33.767(-0.025,-0.001,-0.001) |
29 | 32.448,32.616,32.626 | 32.432,32.591,32.597(-0.016,-0.025,-0.029) |
30 | 31.577,32.609,32.638 | 31.589,32.586,32.610(0.011,-0.023,-0.028) |
31 | 30.827,31.631,31.638 | 30.825,31.631,31.638(-0.003,0.000,0.000) |
32 | 29.643,30.845,30.868 | 29.642,30.845,30.868(-0.001,0.000,0.000) |
33 | 28.581,29.740,29.699 | 28.585,29.740,29.699(0.004,0.000,0.000) |
34 | 27.747,29.733,29.703 | 27.763,29.733,29.703(0.016,0.000,-0.000) |
35 | 26.561,28.759,28.740 | 26.558,28.759,28.740(-0.002,0.000,0.000) |
36 | 25.525,27.850,27.867 | 25.528,27.850,27.867(0.004,0.000,0.000) |
37 | 24.773,27.857,27.857 | 24.784,27.857,27.857(0.012,0.000,0.000) |
38 | 23.640,26.742,26.743 | 23.651,26.729,26.734(0.011,-0.013,-0.009) |
39 | 22.654,26.737,26.736 | 22.653,26.723,26.726(-0.001,-0.013,-0.010) |
40 | 21.787,25.859,25.851 | 21.790,25.859,25.851(0.003,0.000,0.000) |
41 | 20.582,25.859,25.853 | 20.551,25.859,25.853(-0.031,0.000,0.000) |
42 | 19.626,25.203,25.207 | 19.627,25.207,25.212(0.001,0.004,0.005) |
43 | 18.837,25.216,25.211 | 18.840,25.220,25.215(0.003,0.004,0.004) |
44 | 17.653,25.196,25.221 | 17.656,25.200,25.226(0.003,0.004,0.005) |
45 | 16.699,24.358,24.371 | 16.696,24.359,24.372(-0.003,0.001,0.001) |
46 | 15.818,24.362,24.349 | 15.810,24.363,24.350(-0.009,0.001,0.001) |
47 | 14.711,24.368,24.343 | 14.694,24.369,24.344(-0.017,0.001,0.001) |
48 | 13.846,23.798,23.796 | 13.845,23.798,23.795(-0.002,-0.000,-0.001) |
49 | 13.188,23.801,23.785 | 13.199,23.801,23.785(0.011,-0.001,-0.001) |
50 | 12.315,23.789,23.790 | 12.315,23.788,23.789(-0.001,-0.001,-0.001) |
51 | 11.753,23.807,23.792 | 11.758,23.806,23.791(0.005,-0.001,-0.000) |
表3.模拟结果,INTRA16×16编码的宏块,使用α=1/3
如表1-3所示,对于所有允许的量化比例(0-51)和宏块类型(INTER、INTRA4×4或INTRA 16×16),优选实施例简化的正量化几乎同等地执行当前由H.264推荐的量化。因此,优选实施例量化提供与当前H.264量化设计相同的压缩效率,但使得能在不具有32位存储器存取的能力的装置上实施H.264量化。
在保持多个量化表的限制舍入控制参数的位大小的特征的情况下,可对优选实施例作出各种修改。举例来说,量化可使用较精细的解析度,例如增量是qP/8而不是qP/6等。
Claims (5)
1.一种视频编码的方法,其包括以下步骤:
(a)将4×4的整数数据块变换为4×4的整数变换系数块;以及
(b)通过(i)将所述系数的绝对值与多个4×4的正整数量化矩阵中的一者的条目逐元素相乘,(ii)加上舍入控制参数,(iii)恢复所述系数的符号,(iv)且右移,来量化所述4×4的系数块;
(c)其中所述多个量化矩阵中的第一矩阵所具有等于所述多个量化矩阵中的第二矩阵的相应条目二分之一并进行舍入的条目;以及
(d)所述多个量化矩阵中的所述一者是根据量化参数而选出的。
2.根据权利要求1所述的方法,其中:
(a)所述多个量化矩阵包含4×4矩阵M0、M1、...,MQ-1,其中Q是正整数,用因子将其表示为Q=NM,其中N和M是各大于1的正整数;且
(b)对于每对整数n,k,其中n在1到N-1的范围内且k在0到M-1的范围内,所述矩阵的元素的关系为
MnM+k(i,j)=(Mk(i,j)+2n-1)>>n其中0≤i,j≤3。
3.根据权利要求2所述的方法,其中M=6且N=7。
4.根据权利要求2所述的方法,其中:
(a)在所述量化参数等于nM+k,其中n大于0,且所述系数表示为y(i,j)的情况下,所述量化包含以下计算:
c(i,j)=sign[y(i,j)][|y(i,j)|M(n-1)M+k(i,j)+α216]>>16
其中α是舍入因子,且0<α<1。
5.根据权利要求2所述的方法,其中:
(a)在所述量化参数等于k,且所述系数表示为y(i,j)的情况下,所述量化包含以下计算:
c(i,j)=sign[y(i,j)][|y(i,j)|Mk(i,j)+α215]>>15
其中α是舍入因子,且0<α<1。
Applications Claiming Priority (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US65135805P | 2005-02-08 | 2005-02-08 | |
US60/651,358 | 2005-02-08 | ||
US11/347,338 | 2006-02-03 |
Publications (1)
Publication Number | Publication Date |
---|---|
CN101194510A true CN101194510A (zh) | 2008-06-04 |
Family
ID=39488326
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CNA200680010777XA Pending CN101194510A (zh) | 2005-02-08 | 2006-02-08 | H.264量化 |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN101194510A (zh) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103096052A (zh) * | 2011-11-04 | 2013-05-08 | 华为技术有限公司 | 一种图像编码、解码的方法和装置 |
-
2006
- 2006-02-08 CN CNA200680010777XA patent/CN101194510A/zh active Pending
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103096052A (zh) * | 2011-11-04 | 2013-05-08 | 华为技术有限公司 | 一种图像编码、解码的方法和装置 |
WO2013064109A1 (zh) * | 2011-11-04 | 2013-05-10 | 华为技术有限公司 | 一种图像编码、解码的方法和装置 |
CN103096052B (zh) * | 2011-11-04 | 2015-11-25 | 华为技术有限公司 | 一种图像编码、解码的方法和装置 |
US9667958B2 (en) | 2011-11-04 | 2017-05-30 | Huawei Technologies Co., Ltd. | Image coding and decoding methods and apparatuses |
US10091531B2 (en) | 2011-11-04 | 2018-10-02 | Huawei Technologies Co., Ltd. | Image coding and decoding methods and apparatuses |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US7778327B2 (en) | H.264 quantization | |
CN101399988B (zh) | 减少位深的量子化方法 | |
Chen et al. | Video compression using integer DCT | |
JP4153425B2 (ja) | 変換及びそれに続く量子化の実行 | |
EP1113673B1 (en) | Quantization matrix for still and moving picture decoding | |
US20050281332A1 (en) | Transform coefficient decoding | |
CN100412907C (zh) | 视频编码的低复杂性和统一标准的变换 | |
EP1359546A1 (en) | 2-D transforms for image and video coding | |
US7634148B2 (en) | Image signal transforming and inverse-transforming method and computer program product with pre-encoding filtering features | |
WO1999057911A1 (en) | Method and apparatus for inverse quantization of mpeg-4 video | |
CA2248664C (en) | Method and apparatus for fast image compression | |
Varma et al. | JPEG2000-choices and tradeoffs for encoders | |
Ishwar et al. | Discrete tchebichef transform-a fast 4x4 algorithm and its application in image/video compression | |
US20050111554A1 (en) | 8x8 Transform and quantization | |
CN101194510A (zh) | H.264量化 | |
CN100440978C (zh) | 一种视频图像编码方法 | |
Richter | Lossless coding extensions for JPEG | |
CN100411442C (zh) | 一种视频图象编码方法 | |
Lienhart et al. | An FPGA-based video compressor for H. 263 compatible bit streams | |
CN100426866C (zh) | 全相位沃尔什双正交变换及其对jpeg的改进方法 | |
WO2001050767A2 (en) | Method and apparatus for digital video compression using three-dimensional cellular automata transforms | |
Bracamonte et al. | A multiplierless implementation scheme for the JPEG image coding algorithm | |
CN113068033B (zh) | 一种多媒体的反量化处理方法、装置、设备及存储介质 | |
CN100563341C (zh) | 图像和视频编码中多尺度兼容处理装置及其方法 | |
Dolinar et al. | Compressed/reconstructed test images for CRAF/Cassini |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
WD01 | Invention patent application deemed withdrawn after publication |