WO2014032625A1 - 低复杂度的分像素插值滤波方法、滤波装置及电子装置 - Google Patents
低复杂度的分像素插值滤波方法、滤波装置及电子装置 Download PDFInfo
- Publication number
- WO2014032625A1 WO2014032625A1 PCT/CN2013/082789 CN2013082789W WO2014032625A1 WO 2014032625 A1 WO2014032625 A1 WO 2014032625A1 CN 2013082789 W CN2013082789 W CN 2013082789W WO 2014032625 A1 WO2014032625 A1 WO 2014032625A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- pixel
- sub
- interpolation
- filter
- tap
- 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.)
- Ceased
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
- H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
- H04N19/503—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving temporal prediction
- H04N19/51—Motion estimation or motion compensation
- H04N19/523—Motion estimation or motion compensation with sub-pixel accuracy
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
- G06T3/4007—Scaling of whole images or parts thereof, e.g. expanding or contracting based on interpolation, e.g. bilinear interpolation
-
- 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
- H04N19/59—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving spatial sub-sampling or interpolation, e.g. alteration of picture size or resolution
-
- 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
- H04N19/60—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding
- H04N19/61—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding in combination with predictive coding
- H04N19/615—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding in combination with predictive coding using motion compensated temporal filtering [MCTF]
-
- 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
- H04N19/80—Details of filtering operations specially adapted for video compression, e.g. for pixel interpolation
-
- 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
- H04N19/85—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using pre-processing or post-processing specially adapted for video compression
- H04N19/86—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using pre-processing or post-processing specially adapted for video compression involving reduction of coding artifacts, e.g. of blockiness
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N7/00—Television systems
- H04N7/01—Conversion of standards, e.g. involving analogue television standards or digital television standards processed at pixel level
- H04N7/0135—Conversion of standards, e.g. involving analogue television standards or digital television standards processed at pixel level involving interpolation processes
Definitions
- the present invention relates to the field of new generation video coding and decoding technologies, and in particular, to a subpixel interpolation filter and an interpolation method.
- the sub-pixel interpolation technique can improve the accuracy of motion prediction, and thus has a very positive effect on the improvement of the overall codec performance.
- the sub-pixel interpolation technique was first introduced.
- 264/AVC uses a 6-tap filter to obtain 1/2 pixel, and then uses the averaging filter to get 1/4 pixel, the performance has been greatly improved; also in AVS J izhun Prof i le A 4-tap filter is used to obtain 1/2 pixel, and then 1/4 pixel; the next generation of high-performance video codec standard being developed by JCT-VC (the Joint Col labor ive Team on Video Coding) In HEVC (High Efficient Video Coding), a high-performance separable DCT-based interpolation filter is used to further improve the interpolation performance. Compared with H.264/AVC, the bit rate is reduced. Improve coding performance, but the complexity of its interpolation calculation is higher.
- the main object of the present invention is to provide a sub-pixel interpolation filtering method with low complexity and performance balance.
- the present invention proposes a low-complexity sub-pixel interpolation filtering method, which uses an 8-tap interpolation filter and a 6-tap interpolation filter, and uses an 8-tap interpolation filter for adjacent integer pixels to obtain a horizontal or vertical direction. Dividing pixels between adjacent integer pixels; using a horizontal 8-tap interpolation filter between the adjacent integer pixels, and then using a vertical 6-tap interpolation filter, twice Interpolate to get the remaining 9 sub-pixels.
- the present invention also proposes a low complexity sub-pixel interpolation filtering device, which implements video image processing by using the above-described sub-pixel interpolation filtering method.
- the present invention proposes an electronic data carrier storing a computer program for implementing the above-described subpixel interpolation filtering method.
- the present invention proposes an electronic device that processes video images using the low complexity sub-pixel interpolation filtering method described above.
- the invention has the following advantages:
- the sub-pixel interpolation filtering method of the present invention adopts a combination of 8-tap and 6-tap interpolation filters, and the motion variation in the general scene is mainly that the lateral motion is mostly and the longitudinal motion is relatively small, and the 10-tap interpolation filter is considered. It is more complicated, and the 4-tap interpolation filter has poor performance. Therefore, the combination of "8+6" filtering can reduce the computational complexity without significantly affecting performance.
- the computational complexity is slightly higher than the 4-tap filter and much lower than the 10-tap filter, but the performance is very close to the 10-tap filter.
- interpolation filtering method of the present invention optimizes the filter coefficients and uses the best performing filter coefficients.
- the present invention is compared with the luminance interpolation of HEVC, and the code rates of ⁇ , U, and V are only increased by 0.7%, 0.2%, and 0.3%, respectively, but the computational complexity is lower than that of HEVC. 10% (small 9.07%, 6.42%, and 9.36% in terms of number of accesses, multiplications, and additions), with low complexity and while maintaining comparable performance to HEVC luma interpolation.
- FIG. 1 is a schematic diagram of an interpolation method of the present invention
- FIG. 2 is a schematic diagram of an actual interpolation process of the present invention.
- Embodiment 1
- the low-complexity sub-pixel interpolation filtering device of this example adopts the following sub-pixel interpolation filtering method, and the interpolation process is as shown in FIG. 1 , and the position A represented by uppercase letters. ,. , A. 1 , A ⁇ A U is a known integer pixel point, and the lowercase letter represents a sub-pixel point that needs to be interpolated.
- Interpolation filtering is performed using an 8-tap interpolation filter to obtain a sub-pixel a. ,. , b. ,. , c. ,. ;
- the AM uses an 8-tap interpolation filter for interpolation filtering to obtain a sub-pixel d. ,. , h. ,. , n. ,.
- the 8-pixel interpolation filter is used to interpolate the entire pixel of the horizontal nearest neighbor to obtain the intermediate value, and then the intermediate value is interpolated and filtered using the vertical 6-tap interpolation filter.
- the coefficients of the 6-tap and 8-tap interpolation filters are shown in Tables 1, 2, respectively.
- the optimized coefficients in Tables 1 and 2 are for which some of the coefficients try +/- 1 (ie, plus/minus 1), and then based on the experimental results, the optimized coefficients are selected.
- the coefficient obtained according to the algorithm is a decimal number. Generally, these decimals are expanded by 64 times, that is, 6b it is used to represent a coefficient, but the total coefficient sums up and should be 64 (ie, One)). When the coefficient is expanded by 64, there are factors such as intermediate data or rounding. For example, if a coefficient is multiplied by 64 and is 17.51, then 17 and 18 can be tried to see which coefficient performs better.
- the first step According to the coordinates of X and y, the sub-pixel points to be interpolated are determined according to Table 3.
- the second step from the judgment of the previous step, respectively enter the corresponding sub-pixel interpolation processing process: For example, for the sub-pixel a, only the horizontal 8-tap interpolation filtering needs to be performed, and the filter coefficient corresponding to the 1/4 position is used; Pixel e, first perform horizontal 8-tap interpolation filtering, use filter coefficients corresponding to 1 / 4 position, and then perform longitudinal 6-tap interpolation filtering, using filter coefficients corresponding to 1 / 4 position.
- the interpolation process for each specific point is as follows:
- Subpixel a. ,. , b. ,. , 0. ,. Interpolation Interpolation filtering is performed on the entire pixel of the horizontal nearest neighbor using an 8-tap interpolation filter, using filters corresponding to 1/4, 2/4, 3/4 positions, respectively. The coefficient is calculated to obtain the corresponding sub-pixel a. ,. , b. ,. , c. ,.
- Interpolation filtering is performed on the entire pixel of the vertical nearest neighbor using an 8-tap interpolation filter, and the corresponding sub-pixel d is calculated by using the filter coefficients corresponding to the 1/4, 2/4, and 3/4 positions, respectively. . , h. . , n. . , Calculated as follows:
- a'. ;1 differs from a M in that a, M has no final shiftl shift operation; specifically, a, . ,. Can be a. ,. Obtained, the other intermediate values of a, M can be obtained by a M , that is, the intermediate value of a, M can be obtained by a M , a, . , 2 intermediate values can be a. , 2 get, a, . , 3 intermediate values can be a. , 3 get; b, . ; 1 and 6, M is similar.
- the vertical value 6 tap interpolation filtering is used for the intermediate value a' M , and the corresponding sub-pixel e is calculated by using the filter coefficients corresponding to the 1/4, 2/4, 3/4 positions respectively. ,. , i. ,. , ⁇ , ⁇ , the formula is as follows:
- Subpixel g. ,. , k. ,. , ! ⁇ . ,. Interpolation: Firstly, the 8-pixel interpolation filter is used for interpolation of the entire pixel of the horizontal nearest neighbor, and the filter coefficient corresponding to the 3/4 position is used. To the intermediate value c, M (where i -3 ⁇ 4); then to the intermediate value c, . ⁇ Using vertical 6-tap interpolation filtering, using the filter coefficients corresponding to 1/4, 2/4, 3/4 positions, respectively, the corresponding sub-pixel g is calculated. ,. , k. ,. , r. ,. , Calculated as follows:
- the invention effectively reduces the complexity of the interpolation calculation while ensuring the interpolation of higher performance.
- Table 6 4-tap filter coefficient Filter coefficient Multiplication number Force port method
- the 4-tap interpolation filter is compared with the HEVC luminance interpolation.
- the performance is shown in Table 11.
- the U and V code rates are increased by 5.6%, 3.6%, and 3.6%, respectively.
- the 10-tap interpolation filter is compared with the HEVC luminance interpolation.
- the performance is shown in Table 12 below.
- the u and V code rates are only increased by 0.8%, 0.3%, and 0.4%, respectively.
- Embodiment 2 is a diagrammatic representation of Embodiment 1:
- the present example is a computer readable electronic data carrier including, but not limited to, an optical disk, a flash disk and a hard disk, etc., in which a computer program for implementing the low complexity subpixel interpolation filtering device of the first embodiment is stored.
- Embodiment 3 This example is an electronic device in which a program for implementing the low-complexity sub-pixel interpolation filtering method in the first embodiment is built in. By enabling the program, the video image can be processed, specifically, by using an integer for the video image. Interpolation of pixels to obtain sub-pixel points.
- the electronic device uses the low complexity sub-pixel interpolation filtering method in the first embodiment to process the video image.
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Television Systems (AREA)
- Image Processing (AREA)
Description
低复杂度的分像素插值滤波方法、 滤波装置及电子装置 【技术领域】
本发明涉及新一代视频编解码技术领域, 特别涉及一种分像素插值滤 波器及插值方法。
【背景技术】
随着众多如数字电视、 互联网高清视频、 数码相机、 数码摄像机等高 清数码产品的逐渐普及, 现有的视频编解码标准已经不能满足要求, 尤其 对于高分辨率视频, 如 4Kx2K的甚高分辨率。
在视频编解码技术中, 分像素插值技术能提高运动预测的准确度, 从 而对整个编解码性能的提高有非常大的积极作用。 分像素插值技术最早在
MPEG-1中就已经使用, 但仅限于半像素运动估计, 尽管比整像素运动估计 要好, 但是精度和性能还不高; 从 MPEG-4 ASP开始提高了插值精度, 使用 1/4像素精度的插值; H. 264/AVC则采用了一个 6抽头滤波器来获得 1/2 像素, 然后用平均滤波器来得到 1/4 像素, 性能有了很大提高; 在 AVS J izhun Prof i le中也是采用了一个 4抽头滤波器来获得 1/2像素, 然后再 得出 1/4像素; 在最近 JCT- VC (the Joint Col laborat ive Team on Video Coding)正在制订的下一代高性能视频编解码标准 HEVC (High Eff ic i ency Video Cod ing)中, 为进一步提高插值性能, 采用了高性能可分离式的基于 DCT的插值滤波器, 相比于 H. 264/AVC降低了很多码率, 能明显提高编码 性能, 但是其插值计算的复杂度较高。
【发明内容】
本发明的主要目的是: 提供一种低复杂度和性能均衡的分像素插值滤 波方法。
为此, 本发明提出了一种低复杂度的分像素插值滤波方法, 采用 8抽头 插值滤波器和 6抽头插值滤波器, 对于相邻整像素点使用 8抽头插值滤波器, 得到横向或纵向的相邻整像素点之间分像素点; 于所述相邻整像素点之间 分像素点之间, 使用横向的 8抽头插值滤波器, 然后再使用纵向的 6抽头插 值滤波器, 进行两次插值, 得到余下的 9个分像素。
同时, 本发明还提出了一种低复杂度的分像素插值滤波装置, 采用上 述的分像素插值滤波方法实现视频图像处理。
进一步地, 本发明提出了一种电子数据载体, 存储有用于实现上述的 分像素插值滤波方法的计算机程序。
最后, 本发明提出了一种电子装置, 采用上述的低复杂度的分像素插 值滤波方法, 对视频图像进行处理。 本发明具有以下优点:
本发明的分像素插值滤波方法, 采用 8抽头和 6抽头插值滤波器相结 合的方式, 利用一般场景的运动变化主要是横向运动居多而纵向运动相对 较少,且考虑到 10抽头插值滤波器计算较复杂, 而 4抽头插值滤波器又性 能较差, 故采用 "8+6"相结合的滤波方式, 能减小计算复杂度而又不会明 显影响性能。
采用本发明的方法, 其计算的复杂度略高于 4抽头滤波器, 同时大大 低于 10抽头滤波器, 但性能却非常接近 10抽头滤波器。
进一步地, 本发明的插值滤波方法, 并对滤波器系数进行了优化, 采 用了性能最好的滤波器系数。
具体的, 本发明与 HEVC的亮度插值对比, 其¥、 U、 V码率分别仅增加 了 0. 7%、 0. 2%、 0. 3%; 但是其计算复杂度, 比 HEVC要降低近 10% (在访存 数量、乘法和加法数目方面分别小 9. 07%、 6. 42%和 9. 36% ), 具有低复杂度 的优势且同时保持了与 HEVC亮度插值有相当的性能。
【附图说明】
图 1是本发明的插值方法示意图;
图 2是本发明的实际插值过程示意图。
【具体实施方式】
下面通过具体的实施例并结合附图对本发明作进一步详细的描述。 实施例一:
本例的低复杂度的分像素插值滤波装置, 采用如下的分像素插值滤波 方法, 其插值过程如图 1所示, 大写字母代表的位置 A。,。、 A。; 1、 A ^ AU 为已知的整像素点, 小写字母代表的位置为需要插值得到的分像素点。 对 横向最近邻的整像素点 A。,。、 A1;。使用 8抽头插值滤波器进行插值滤波, 得 到分像素 a。,。、 b。,。、 c。,。; 对纵向最近邻的整像素点 A。,。、 AM使用 8抽头插 值滤波器进行插值滤波, 得到分像素 d。,。、 h。,。、 n。,。; 对于剩下的分像素是 先对横向最近邻的整像素点使用 8抽头插值滤波器进行插值滤波, 得到中 间值, 然后再对中间值使用纵向 6抽头插值滤波器进行插值滤波。 6抽头 和 8抽头插值滤波器的系数分别见表 1, 2所示。
表 1和表 2中经优化的系数是对其中部分系数尝试 +/-1 (即加 /减 1 ), 然后根据实验结果, 选用优化的系数。 各系数可以加减 1的原因是: 根据 算法得到的系数是个小数, 一般是对这些小数进行扩大 64倍, 即用 6b i t 来表示一个系数, 但是总的系数加起来和应为 64 (即归一化)。 当对每个 系数扩大 64时, 由于存在中间数据或四舍五入等因素, 比如某个系数乘以 64后是 17. 51 , 此时就可以尝试 17和 18 , 看哪个系数性能更好。
表 1 8抽头插值滤波器的系数列表
第一步:根据表 3所示,通过 X和 y的坐标来判断所要插值的分像素点。
表 3 分像素点与坐标位置对应关系表
第二步: 由上一步的判断, 分别进入相应的分像素插值处理过程: 如, 对于分像素 a , 只需要进行横向的 8抽头插值滤波, 使用 1 /4位置对应的 滤波器系数; 对于分像素 e , 先进行横向的 8抽头插值滤波, 使用 1 /4位 置对应的滤波器系数, 然后再进行纵向的 6抽头插值滤波, 使用 1 /4位置 对应的滤波器系数。 具体各个点的插值过程如下:
(1) 分像素 a。,。、 b。,。、 0。,。的插值: 对横向最近邻的整像素点使用 8抽 头插值滤波器进行插值滤波, 分别使用 1 /4 , 2/4 , 3/4位置对应的滤波器
系数, 计算得到对应的分像素 a。,。、 b。,。、 c。,。, 计算公式如下: a00 = (-Λ + 4 χ Λ_20 -10χΛ_10 +57x +18χΛ10 -6> ^0 + 3χ ^,0 - Λ40 ) » shiftl b0,o = (—A— 3,0 +4x A_2i0— 11 x A_10 +40x 0 +40xA10 -llx A2>0 + 4 x — A4>0 ) » shiftl c0,0 =(-A_30 +3xA_20 -6xA_10 +18x ),o +57 x^0 -lOx^o +4x -A40)» shiftl (2) 分像素 d。。、 h。。、 n。。的插值: 对纵向最近邻的整像素点使用 8抽头 插值滤波器进行插值滤波, 分别使用 1/4, 2/4, 3/4位置对应的滤波器系 数, 计算得到对应的分像素 d。。、 h。。、 n。。, 计算公式如下:
d00 = (-
) » shiftl h„,„ = (-Λ +4 A0>_2— 11 x A,,,-! + 40 x ,0 + 40 x A 11 x +4 A0>3 - A ) » shiftl n0,0 = (- _3 + 3 x _2 - 6x^i0_1 +18x +57x^ ^ -10x^„2 +4x 4) » shiftl
(3) 分像素 e。,。、 i。,。、 p。,。的插值: 先对横向最近邻的整像素点使用 8 抽头插值滤波, 使用 1/4位置对应的滤波器系数, 从而得到中间值 a, 。
(其中 i=_3~4), a'。;1与 aM区别在于 a, M没有最后的 shiftl移位操作; 具体地, a, 。,。可由 a。,。得到, 同理其他的 a, M中间值可由 aM得到, 即 a, M中间值可由 aM得到, a, 。,2中间值可由 a。, 2得到, a, 。,3中间值可由 a。,3 得到; b, 。;1和6, M类似。
然后再对中间值 a' M使用纵向 6抽头插值滤波, 分别使用 1/4, 2/4, 3/4位置对应的滤波器系数, 计算得到对应的分像素 e。,。、 i。,。、 ρο,ο, 计算 公式如下:
e0,。 = (2 a'0,— 2 - 9 a'0,— ' +57x a'00 +17 a'01 - 4 a'0,2 + a'03 ) » shiftl i00 = (2 x a'0 _2 - 9 x a'0 _x + 39 x a'00 + 39 x a'01 - 9 x a'02 + 2x a'03 ) » shiftl Ρο,ο = (a'o,— -4xa'0 j +17xa'00 +57xa'01 - 9xa'02 + 2xa'0,3) >> shiftl
(4) 分像素 f。,。、 j。,。、 ¾。的插值: 先对横向最近邻的整像素点使用 8 抽头插值滤波, 使用 2/4位置对应的滤波器系数, 从而得到中间值 b, 。 (其中 i=-3~4); 然后再对中间值 b, M使用纵向 6抽头插值滤波, 分别 使用 1/4, 2/4, 3/4位置对应的滤波器系数, 计算得到对应的分像素 f。,。、 j。,。、 ¾。, 计算公式如下:
f0。 =(2xb'0— 2 _9xb'0 + 57x '0。 + 17xb'01_4xb'o 2 + b'0 >> shiftl
j。。 = (2x b 'o 2_ 9 x b Ό— 39 x b Ό。+ 39 x b 'o「 9 x b 'o 2 + 2x b Ό 3 ) >> shift 2 q0,0 = (b Ό - 4 x b + 17xb'00 + 57xb '01 -9xb '02 +2xb '03 ) » shiftl
(5) 分像素 g。,。、 k。,。、 !~。,。的插值: 先对横向最近邻的整像素点使用 8 抽头插值滤波器进行插值滤波, 使用 3/4位置对应的滤波器系数, 从而得
到中间值 c, M (其中 i=-3~4); 然后再对中间值 c, 。^使用纵向 6抽头插 值滤波, 分别使用 1/4, 2/4, 3/4位置对应的滤波器系数, 计算得到对应 的分像素 g。,。、 k。,。、 r。,。, 计算公式如下:
g。。=(2xc'0— 2_9xc'0— i + S xc'oo + nxc'ofAxc'oz+c'o shift!
k。。 = (2 x c '0— 2 _ 9 x c ' 丄 + 39 x c 'o。 + 39 x c 'o i _ 9 x c '02 + 2 x c '03 ) >> shift! r0。 =(c'0— 2_4xc'0— i + nxc'oo + S xc'ofQxc'oz + Sxc'o » shift!
上述公式中, shiftl=6, shift2=12; shift 值的选取依据是: 要使最 终的结果仍保持一个像素的大小即 8bit, 比如, 要插值一个像素, 那么先 横向插值 (乘以横向插值滤波器的各个系数, 这些系数加起来一共等于 64 ), 则该像素被扩大了 64倍, 即相当于左移了 6位, 然后进行纵向插值, 同样的又被左移了 6位,所以最后需要再右移 6+6=12位才恢复一个像素的 大小。 如果只有一次插值过程, 则最后需要右移 6位, 如果有两次插值过 程则需要移 12位。
这样, 所有的 15个分像素点都被插值出来。 至此, 该实施例的所有步 骤完成。
该发明通过以上的插值方法, 在保证了较高性能的插值的同时也有效 的减小了插值计算的复杂度。
( 1 )本例与 HEVC的亮度插值对比,其 Y、U、V码率分别仅增加了 0.7%、 0.2%、 0.3%, 见表 4; 但是其计算复杂度, 如表 5所示, 比 HEVC要降低近 10% (在访存数量、 乘法和加法数目方面分别小 9.07%、 6.42%和 9.36%), 具有低复杂度的优势, 又同时保持了与 HEVC亮度插值有相当的性能。 表 4 本例的方法与 HEVC亮度插值方法的码率对比结果表
分像 算术操作 算术操作 素 访存 乘法 力口法 访存 乘法 力口法 数 数 数 数
A 1 0 0 1 0 0 a 8 6 7 7 5 6 b 8 6 7 8 6 7 c 8 6 7 7 5 6 d 8 6 7 7 5 6 e 8*6 6*6+5 7*6+5 7*7 5*7+5 6*7+6 f 8*6 6*6+5 7*6+5 7*8 6*7+5 7*7+6 g 8*6 6*6+5 7*6+5 7*7 5*7+5 6*7+6 h 8 6 7 8 8 7 i 8*6 6*6+6 7*6+5 8*7 5*8+6 6*8+7 j 8*6 6*6+6 7*6+5 8*8 6*8+6 7*8+7 k 8*6 6*6+6 7*6+5 8*7 5*8+6 6*8+7 n 8 6 7 7 7 6
P 8*6 6*6+5 7*6+5 7*7 5*7+5 6*7+6 q 8*6 6*6+5 7*6+5 7*8 6*7+5 7*7+6 r 8*6 6*6+5 7*6+5 7*7 7*7+5 6*7+6 avg 30. 06 29. 06 33. 06 27. 25 32. 06
( 2 )采用本例的方法, 与采用 4抽头或 10抽头插值滤波器相比, 可 以分析两者的参考像素的数目, 用到的加法数目和乘法数目三个方面, 具 体数据对比如下。
表 6 4抽头滤波器系数 滤波器系数 乘法数 力口法数
1/4 (- -6, 56, 15, - -1} 3 3
2/4 (- -4, 36, 36, - -4} 4 3
3/4 (- -1, 15, 56, - -6} 3 3
表 7 10抽头滤波器系数
位置 滤波器系数 乘法数 力口法数
1/4 U, -2, 4, -10, 57, 19, -7, 3, - 1, 0} 7 8
2/4 U, -2, 5, -12, 40, 40, - -12, 5, -2, 1} 8 9
3/4 {0, -1, 3, -7, 19, 57, - 10, 4, -2, 1} 7 8 对比结果如下表 8、 表 9和表 10所示, 可见, 采用本例的方法, 其计 算的复杂度略高于 4抽头滤波器, 同时大大低于 10抽头滤波器,但性能却 接近于 10抽头滤波器。
表 8 最大像素访存数
Mults ( x Wx L)
Pel 4抽头 10抽头 本发明
A 0 0 0 a 3 7 6 b 4 8 6 c 3 7 6 d 3 7 6 e 3 x 4+3 7 x 10+7 6 x 6+5
f 4 x 4+3 8 x 10+7 6 x 6+5
g 3 x 4+3 7 x 10+7 6 x 6+5
/ 68Ζ80Π3/:0ΖΖ1><1s OlozM
( 3 )采用 4抽头插值滤波器与 HEVC的亮度插值对比,性能如下表 11, U、 V码率分别增加了 5.6%、 3.6%、 3.6%。
表 11
(4)采用 10抽头插值滤波器与 HEVC的亮度插值对比, 性能如下表 12, u、 V码率仅分别增加了 0.8%、 0.3%、 0.4%。
表 12
实施例二:
本领域技术人员可以理解, 上述实施方式中各种方法的全部或部分步 骤可以通过计算机程序来指令相关硬件完成。 本例是一种计算机可读的电 子数据载体, 包括但不限于光盘、 闪存盘和硬盘等, 其中存储了用于实现 实施例一中的低复杂度的分像素插值滤波装置的计算机程序。 实施例三:
本例是一种电子装置, 其中内置了实现实施例一中的低复杂度的分像 素插值滤波方法的程序, 通过启用该程序, 可对视频图像进行处理, 具体 的, 即通过对视频图像整数像素点的插值, 得到分像素点。 或者, 该电子 装置, 采用了实施例一中的低复杂度的分像素插值滤波方法, 对视频图像 进行处理。 以上内容是结合具体的优选实施方式对本发明所作的进一步详细说 明, 不能认定本发明的具体实施只局限于这些说明。 对于本发明所属技术 领域的普通技术人员来说, 在不脱离本发明构思的前提下, 还可以做出若 干筒单推演或替换, 都应当视为属于本发明的保护范围。
Claims
1.一种低复杂度的分像素插值滤波方法, 采用 8抽头插值滤波器和 6抽 头插值滤波器, 对于相邻整像素点使用 8抽头插值滤波器, 得到横向或纵向 的相邻整像素点之间的分像素点; 于所述相邻整像素点之间的分像素点之 间, 使用横向的 8抽头插值滤波器, 然后再使用纵向的 6抽头插值滤波器, 进行两次插值, 得到余下的 9个分像素。
2.如权利要求 1所述的分像素插值滤波方法, 其特征是: 所述 8抽头插 值滤波器的插值系数为: 1/4像素对应的系数为 {-1, 4, -10, 57, 18, _6, 3, -1} , 1/2像素对应的系数为 {-1, 4, -11, 40, 40, -11, 4, -1} , 3/4 像素对应的系数为 {_1, 3, -6, 18, 57, -10, 4, -1}。
3. 如权利要求 2所述的分像素插值滤波方法, 其特征是: 所述 6抽头插 值滤波器的插值系数为: 1/4像素对应的系数为 {2, -9, 57, 17, _4, 1} , 1/2像素对应的系数为 {2, -9, 39, 39, _9, 2} , 3/4像素对应的系数为 {1, -4, 17, 57, -9, 2}。
4.如权利要求 3所述的分像素插值滤波方法,其特征是: 其中的分像素 a。,。、 b。,。、 c。,。的插值过程包括: 对横向最近邻的整像素点使用 8抽头插值 滤波器进行插值滤波, 分别使用 1/4, 2/4, 3/4位置对应的滤波器系数, 计算得到对应的分像素 a。,。、 b。,。、 c。,。, 计算公式如下:
5. 如权利要求 3所述的分像素插值滤波方法,其特征是:其中的分像 素 d。。、 h。。、 n。。的插值过程包括:: 对纵向最近邻的整像素点使用 8抽头插 值滤波器进行插值滤波, 分别使用 1/4, 2/4, 3/4位置对应的滤波器系数, 计算得到对应的分像素 d。。、 h。。、 n。。, 计算公式如下:
6. 如权利要求 3所述的分像素插值滤波方法, 其特征是: 其中的分像 素 e。,。、 i。,。、 p。,。的插值过程包括: 先对横向最近邻的整像素点使用 8抽头 插值滤波, 使用 1/4位置对应的滤波器系数, 从而得到中间值 a, M (其中
i=_3~4);然后再对中间值 a' M使用纵向 6抽头插值滤波,分别使用 1/4, 2/4, 3/4位置对应的滤波器系数, 计算得到对应的分像素 e。,。、 i。,。、 ρο,οο
7. 如权利要求 3所述的分像素插值滤波方法, 其特征是: 其中的分像 素 f。,。、 j。,。、 ¾。的插值过程包括: 先对横向最近邻的整像素点使用 8抽头 插值滤波, 使用 2/4位置对应的滤波器系数, 从而得到中间值 b, M (其中 i=_3~4);然后再对中间值 b' M使用纵向 6抽头插值滤波,分别使用 1/4, 2/4, 3/4位置对应的滤波器系数, 计算得到对应的分像素 f。,。、 j。,。、 ¾。, 计算公式如下:
f0。 =(2xb'0— 2 _9xb'04 + 57x '0。 + 17xb'01_4xb'o 2 + b'0 >> shift!
j。。 = (2x b 'o— 2_ 9 x b Ό— 39 x b Ό。+ 39 x b 'o「 9 x b 'o 2 + 2x b Ό 3 ) >> shift! q。。 = (V0— 2_4xb'0— i + nxb'oo + S xb'ofQxb'oz+ xb'o >> shift!。
8. 如权利要求 3所述的分像素插值滤波方法, 其特征是: 其中的分像 素 g。,。、 k。,。、 r。,。的插值过程包括: 先对横向最近邻的整像素点使用 8抽头 插值滤波器进行插值滤波, 使用 3/4位置对应的滤波器系数, 从而得到中 间值 c, M (其中 i=-3~4); 然后再对中间值 c, M使用纵向 6抽头插值滤 波, 分别使用 1/4, 2/4, 3/4位置对应的滤波器系数, 计算得到对应的分 像素 g。,。、 k。,。、 r。,。, 计算公式: ¾口下:
g。。 =(2xc'0— 2_9xc'0— i + S xc'oo + nxc'ofAxc'oz+c'o shift!
k。。 = (2x c'0 2_ 9x c'0 39x c'0。 + 39x c'0「 9x c'02 + 2x c'03 ) >> shift! r00 = (c'0_2-4xc'0_1 + 17xc'00 + 57xc'01-9xc'02 + 2xc'03) » shift!。
9. 如权利要求 4或 5所述的分像素插值滤波方法, 其特征 : 所述 shiftl的值为 6。
10.如权利要求 6-8中任一项所述的分像素插值滤波方法, 其特征是: 所述 shift2的值为 12。
11.一种低复杂度的分像素插值滤波装置, 采用如权利要求 1-10中任一 项所述的分像素插值滤波方法实现视频图像处理。
12. 一种电子数据载体, 存储有用于实现如权利要求 1-10中任一项所 述的分像素插值滤波方法的计算机程序。
13.一种电子装置, 采用如权利要求 1-10中任一项所述的低复杂度的分 像素插值滤波方法, 对视频图像进行处理。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| US14/636,162 US20150172702A1 (en) | 2012-09-03 | 2015-03-02 | Fractional-pel interpolation filter method, filter device and electronic device using the same |
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210321179.1 | 2012-09-03 | ||
| CN2012103211791A CN102833550A (zh) | 2012-09-03 | 2012-09-03 | 一种低复杂度的分像素插值滤波器 |
Related Child Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| US14/636,162 Continuation-In-Part US20150172702A1 (en) | 2012-09-03 | 2015-03-02 | Fractional-pel interpolation filter method, filter device and electronic device using the same |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2014032625A1 true WO2014032625A1 (zh) | 2014-03-06 |
Family
ID=47336484
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/CN2013/082789 Ceased WO2014032625A1 (zh) | 2012-09-03 | 2013-09-02 | 低复杂度的分像素插值滤波方法、滤波装置及电子装置 |
Country Status (3)
| Country | Link |
|---|---|
| US (1) | US20150172702A1 (zh) |
| CN (1) | CN102833550A (zh) |
| WO (1) | WO2014032625A1 (zh) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114390279A (zh) * | 2022-01-12 | 2022-04-22 | 北京达佳互联信息技术有限公司 | 亚像素插值方法和装置 |
Families Citing this family (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101682147B1 (ko) * | 2010-04-05 | 2016-12-05 | 삼성전자주식회사 | 변환 및 역변환에 기초한 보간 방법 및 장치 |
| CN102833550A (zh) * | 2012-09-03 | 2012-12-19 | 北京大学深圳研究生院 | 一种低复杂度的分像素插值滤波器 |
| CN103974078B (zh) * | 2014-03-31 | 2017-10-31 | 北京大学深圳研究生院 | 一种针对色度的插值方法及滤波器 |
| WO2015149241A1 (zh) * | 2014-03-31 | 2015-10-08 | 北京大学深圳研究生院 | 一种针对色度的插值方法及滤波器 |
| US10462480B2 (en) * | 2014-12-31 | 2019-10-29 | Microsoft Technology Licensing, Llc | Computationally efficient motion estimation |
| CN106331702A (zh) * | 2015-06-30 | 2017-01-11 | 阿里巴巴集团控股有限公司 | 一种分像素插值滤波方法及设备 |
| CN105791876B (zh) * | 2016-03-14 | 2018-11-13 | 杭州电子科技大学 | 基于低复杂度分层插值的hevc分像素运动估计方法 |
| CN106507118B (zh) * | 2016-11-28 | 2019-10-11 | 浪潮集团有限公司 | 一种双模亮度插值滤波结构及方法 |
| CN116389743A (zh) * | 2023-04-18 | 2023-07-04 | 哲库科技(上海)有限公司 | 图像编码方法、装置、芯片、终端、存储介质及程序产品 |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1543221A (zh) * | 2003-11-04 | 2004-11-03 | 浙江大学 | 1/4像素精度插值的方法和装置 |
| WO2012100085A1 (en) * | 2011-01-19 | 2012-07-26 | General Instrument Corporation | High efficiency low complexity interpolation filters |
| CN102833550A (zh) * | 2012-09-03 | 2012-12-19 | 北京大学深圳研究生院 | 一种低复杂度的分像素插值滤波器 |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101682147B1 (ko) * | 2010-04-05 | 2016-12-05 | 삼성전자주식회사 | 변환 및 역변환에 기초한 보간 방법 및 장치 |
-
2012
- 2012-09-03 CN CN2012103211791A patent/CN102833550A/zh active Pending
-
2013
- 2013-09-02 WO PCT/CN2013/082789 patent/WO2014032625A1/zh not_active Ceased
-
2015
- 2015-03-02 US US14/636,162 patent/US20150172702A1/en not_active Abandoned
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1543221A (zh) * | 2003-11-04 | 2004-11-03 | 浙江大学 | 1/4像素精度插值的方法和装置 |
| WO2012100085A1 (en) * | 2011-01-19 | 2012-07-26 | General Instrument Corporation | High efficiency low complexity interpolation filters |
| CN102833550A (zh) * | 2012-09-03 | 2012-12-19 | 北京大学深圳研究生院 | 一种低复杂度的分像素插值滤波器 |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114390279A (zh) * | 2022-01-12 | 2022-04-22 | 北京达佳互联信息技术有限公司 | 亚像素插值方法和装置 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102833550A (zh) | 2012-12-19 |
| US20150172702A1 (en) | 2015-06-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2014032625A1 (zh) | 低复杂度的分像素插值滤波方法、滤波装置及电子装置 | |
| CN106331722B (zh) | 图像预测方法和相关设备 | |
| US8179446B2 (en) | Video stabilization and reduction of rolling shutter distortion | |
| TWI664856B (zh) | 對影像解碼之方法 | |
| ES2266665T3 (es) | Control del redondeo para la interpolacion multietapa. | |
| CN113491114A (zh) | 用于帧间预测的运动向量舍入、截取和存储的方法和装置 | |
| KR101855975B1 (ko) | 모션 보상 예측을 위한 방법 및 장치 | |
| Parker et al. | Global and locally adaptive warped motion compensation in video compression | |
| BRPI0910641B1 (pt) | técnicas avançadas de interpolação para compensação de movimento em codificação de vídeo | |
| JP2003333603A (ja) | コンピュータ実施される方法、システムおよびコンピュータ可読媒体 | |
| TWI271106B (en) | Apparatus and method for motion estimation supporting multiple video compression standards | |
| CN103747269B (zh) | 一种滤波器插值方法及滤波器 | |
| JP2007515115A (ja) | 内挿される画素値の計算方法の改良 | |
| CN103621089A (zh) | 动态图像编码装置、动态图像解码装置、动态图像编码方法、动态图像解码方法、动态图像编码程序及动态图像解码程序 | |
| WO2021056215A1 (zh) | 运动估计方法、系统和存储介质 | |
| CN100455025C (zh) | 一种分数像素插值方法和编解码设备 | |
| CN116491120A (zh) | 用于仿射运动补偿预测细化的方法和装置 | |
| CN103747251B (zh) | 一种视频图像的插值优化方法及滤波器 | |
| CN112422976B (zh) | 一种视频编码标准中的亮度分量运动补偿方法及视频编码方法 | |
| US9549205B2 (en) | Method and device for encoding video | |
| KR100794098B1 (ko) | 화소연산장치 | |
| CN116171576A (zh) | 用于仿射运动补偿预测细化的方法和装置 | |
| CN103974078B (zh) | 一种针对色度的插值方法及滤波器 | |
| JP2011249986A (ja) | 動き補償装置 | |
| US8422829B2 (en) | Interpolation method and interpolation circuit |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 13833501 Country of ref document: EP Kind code of ref document: A1 |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |
|
| 122 | Ep: pct application non-entry in european phase |
Ref document number: 13833501 Country of ref document: EP Kind code of ref document: A1 |







