CN101986711A - 1/8 pixel precision interpolation method and interpolation device - Google Patents

Abstract

The invention discloses a 1/8 pixel precision interpolation method, comprising the following steps: setting the category of sub-pixel points and the position of the sub-pixel points; selecting the distance from the sub-pixel block formed by the sub-pixel points from the original reference frame Integer pixels in the nearest (M+5)×(N+5) integer pixel block are used as reference pixels, and interpolation is performed according to the positions of various sub-pixels that have been set to obtain the sub-pixels. The present invention only needs to use integer pixels as reference pixels for interpolation at all sub-pixel positions, which simplifies the interpolation process in sub-pixel positions, thereby reducing the time delay of the interpolation process, and can make the motion vector accurate to 1/8 in motion compensation Pixel level, improve the prediction accuracy, reduce the bit rate, and improve the compression efficiency of video coding.

Description

A 1/8 pixel precision interpolation method and interpolation device

technical field

The invention relates to the technical field of video coding, in particular to an interpolation method and an interpolation device with 1/8 pixel precision.

Background technique

Most of today's video coding standards (such as H.261, H.263, H.264, MPEG-1, MPEG-2, MPEG-4, AVS) adopt a hybrid coding framework, and comprehensively use prediction, transformation, Technologies such as entropy coding implement video compression.

In existing video coding technologies, video data is generally divided into intra-frame coding frames and inter-frame coding frames. Generally, in the process of encoding the data in an inter-coded frame, the reference data is first found in the reference frame using motion search technology, and the motion vector is encoded to indicate the position of the reference data in the reference frame at the decoding end. In the early video coding standards, motion compensation with integer pixel precision was adopted, such as H.261. Today's video coding standards adopt motion compensation with sub-pixel precision. MPEG-1 and MPEG-2 introduce motion compensation with 1/2 pixel precision, while H.263, MPEG-4 and H.264 further introduce 1/2 pixel precision motion compensation. Motion compensation with 4-pixel precision. Practice has proved that high-precision motion compensation can improve the compression efficiency of video coding.

For motion compensation with sub-pixel precision, the reference data at the sub-pixel level is obtained by interpolating the reference data of the entire pixel in the reference frame. Early motion compensation with 1/2 pixel precision usually uses bilinear difference. The motion compensation of 1/4 pixel precision in H.264 uses the Wiener filter for interpolation. However, the interpolation process in today's sub-pixel interpolation method is complex, resulting in a large time delay in the interpolation process, which is not conducive to motion compensation, and the compression efficiency of video coding is low.

Contents of the invention

The purpose of the present invention is to at least solve one of the above-mentioned technical defects, and in particular provide an interpolation method with 1/8 pixel precision, which can reduce the time delay of obtaining sub-pixel interpolation process, and is beneficial to motion compensation, reducing bit rate and improve the compression efficiency of video encoding.

Another object of the present invention is to provide an interpolation device with 1/8 pixel precision.

For this reason, the embodiment of the first aspect of the present invention proposes a kind of interpolation method of 1/8 pixel precision, comprises the following steps: setting sub-pixel point category and the position of described sub-pixel point, and described sub-pixel point comprises: pixel, 1/4 pixel and 1/8 pixel, wherein, the half pixel includes a type of half pixel and a second type of half pixel, and the type of half pixel is the same as the integer pixel For the half-pixels in the same row or column, the second type of half-pixels are the remaining half-pixels in the half-pixels;

The 1/4 pixels include one type of 1/4 pixels, two types of 1/4 pixels, and three types of 1/4 pixels, and the one type of 1/4 pixels is the same as the integer pixel A 1/4 pixel in a row or the same column, the second type of 1/4 pixel point is in the same row with the first type of half pixel point and the second type of half pixel point at the same time or simultaneously with the first type of half pixel point and the second type of half pixel point The 1/4 pixel points of the same half-pixel point are in the same column, and the three types of 1/4 pixel points are the remaining 1/4 pixel points in the 1/4 pixel points;

The 1/8 pixel points include class 1/8 pixel points, class 2 1/8 pixel points, class 3 1/8 pixel points, class 4 1/8 pixel points, class 5 1/8 pixel points, class 6 1/8 pixels, seven types of 1/8 pixels, eight types of 1/8 pixels, and nine types of 1/8 pixels, the type 1/8 pixels are located between the integer pixels and the type 1/8 1/8 pixels between 4 pixels; the second type of 1/8 pixels are 1/8 pixels between the first type of half pixels and the first type of 1/4 pixels, and the three Class 1/8 pixel points are 1/8 pixel points located between the first class 1/4 pixel points and third class 1/4 pixel points, and the four class 1/8 pixel points are 1/8 pixel points located in the first class and a half The 1/8 pixel point between the pixel point and the second type 1/4 pixel point, the fifth type 1/8 pixel point is the 1/8 pixel point between the third type 1/4 pixel point and the second type 1/4 pixel point 8 pixels, the six types of 1/8 pixels are 1/8 pixels between the second type of 1/4 pixels and the second type of half pixels, and the seven types of 1/8 pixels are located The 1/8 pixels between the 1/8 pixels of the first category and the 1/8 pixels of the third category, the 1/8 pixels of the eight categories are located in the 1/8 pixels of the second category, the 1/8 pixels of the third category /8 pixels, four types of 1/8 pixels, 1/8 pixels between five types of 1/8 pixels, the nine types of 1/8 pixels are located in the five types of 1/8 pixels and six types of pixels 1/8 pixels between 1/8 pixels; and

Select the integer pixel in the (M+5)×(N+5) integer pixel block closest to the sub-pixel block formed by the sub-pixel from the original reference frame as the reference pixel, according to the set various types The position of the sub-pixels is interpolated to obtain the sub-pixels, wherein M is the number of sub-pixels in each row in the sub-pixel block to be obtained by interpolation, and N is the number of sub-pixels in each row of the sub-pixels to be obtained by interpolation. The number of pixels in the column.

According to the interpolation method with 1/8 pixel precision in the embodiment of the present invention, for all sub-pixel position interpolation, only integer pixels need to be used as reference pixels, which simplifies the interpolation process in the sub-pixel position, thereby reducing the time delay of the interpolation process. At the same time, a high-precision interpolation filter is used, which can make the motion vector accurate to 1/8 pixel level in motion compensation, improve the prediction accuracy, reduce the bit rate, and improve the compression efficiency of video coding.

The embodiment of the second aspect of the present invention proposes a 1/8 pixel precision interpolation device, including: an acquisition module, used to acquire an integer pixel as a reference pixel from the original reference frame; a sub-pixel interpolation module, the The sub-pixel interpolation module is connected with the acquisition module, and is used to receive integer pixels from the acquisition module and interpolate the integer pixels to obtain sub-pixels, and the sub-pixels include: half-pixels, 1/4 pixel point and 1/8 pixel point, the sub-pixel point interpolation module includes: a type of half pixel interpolation unit, used for interpolation to obtain a type of half pixel point, the type of half pixel point is related to the integer half-pixels in the same row or column; a second-type half-pixel interpolation unit for interpolating to obtain second-type half-pixels, the second-type half-pixels being the remaining half-pixels in the half-pixels; A type of 1/4 pixel interpolation unit, used for interpolation to obtain a type of 1/4 pixel, the type of 1/4 pixel is a 1/4 pixel in the same row or column as the integer pixel; two A class 1/4 pixel interpolation unit, used for interpolation to obtain a second class of 1/4 pixels, the second class of 1/4 pixels being in the same row or simultaneously with the first class of half pixels and the second class of half pixels The 1/4 pixel points in the same column as the first-type half-pixel point and the second-type half-pixel point; the third-type 1/4-pixel interpolation unit is used for interpolation to obtain three types of 1/4 pixel points, and the three-type 1/4 pixel points are The /4 pixel point is the remaining 1/4 pixel point within the 1/4 pixel point; a type of 1/8 pixel interpolation unit is used for interpolation to obtain a type of 1/8 pixel point, and the type of 1/8 pixel point The point is a 1/8 pixel point between the integer pixel and a first-class 1/4 pixel point; a second-class 1/8 pixel interpolation unit is used for interpolation to obtain a second-class 1/8 pixel point, and the second-class 1/8 pixel The /8 pixel point is a 1/8 pixel point between the first type of half pixel point and the first type of 1/4 pixel point; the third type of 1/8 pixel interpolation unit is used for interpolation to obtain three types of 1/8 pixel points , the three types of 1/8 pixels are 1/8 pixels located between the first type of 1/4 pixels and the three types of 1/4 pixels; the four types of 1/8 pixel interpolation units are used for interpolation Obtain four types of 1/8 pixels, the four types of 1/8 pixels are 1/8 pixels located between the first type of half pixels and the second type of 1/4 pixels; five types of 1/8 pixels An interpolation unit, configured to interpolate and obtain five types of 1/8 pixel points, the five types of 1/8 pixel points being 1/8 pixel points between the third type of 1/4 pixel points and the second type of 1/4 pixel points; Six types of 1/8 pixel interpolation units, used for interpolation to obtain six types of 1/8 pixel points, the six types of 1/8 pixel points are located between the second type of 1/4 pixel points and the second type of half pixel points 1/8 pixels; seven types of 1/8 pixel interpolation units, used for interpolation to obtain seven types of 1/8 pixels, the seven types of 1/8 pixels are located in the first type of 1/8 pixels and the third type 1/8 pixels between 1/8 pixels; eight types of 1/8 pixel interpolation units, used for interpolation to obtain eight types of 1/8 pixels, the eight types of 1/8 pixels are located in the second type 1/8 pixel, three types 1/ 8 pixels, four types of 1/8 pixels, 1/8 pixels between five types of 1/8 pixels; nine types of 1/8 pixel interpolation unit, used for interpolation to obtain nine types of 1/8 pixels, so The nine types of 1/8 pixels are 1/8 pixels between five types of 1/8 pixels and six types of 1/8 pixels; and an output module, the output module is connected to the sub-pixel interpolation module , used to output the sub-pixel value of the sub-pixel obtained by interpolation.

According to the 1/8 pixel precision interpolation device of the embodiment of the present invention, for all sub-pixel position interpolation, only integer pixels need to be used as reference pixels, which simplifies the interpolation process in sub-pixel positions, thereby reducing the time delay of the interpolation process. At the same time, the interpolation device of the embodiment of the present invention adopts a high-precision interpolation filter, which can make the motion vector accurate to 1/8 pixel level in motion compensation, improve the prediction accuracy, reduce the code rate, and improve the compression efficiency of video coding.

Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.

Description of drawings

The above and/or additional aspects and advantages of the present invention will become apparent and easy to understand from the following description of the embodiments in conjunction with the accompanying drawings, wherein:

Fig. 1 is the flowchart of the interpolation method of 1/8 pixel precision according to the embodiment of the present invention;

2 is a schematic diagram of the distribution of integer pixels used as reference pixels according to an embodiment of the present invention;

3 is a schematic diagram of the distribution of sub-pixels and integer pixels according to an embodiment of the present invention;

4 is a schematic structural diagram of an interpolation device with 1/8 pixel precision according to an embodiment of the present invention; and

FIG. 5 is a schematic structural diagram of the sub-pixel interpolation module in FIG. 4 .

Detailed ways

Embodiments of the present invention are described in detail below, examples of which are shown in the drawings, wherein the same or similar reference numerals designate the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the figures are exemplary only for explaining the present invention and should not be construed as limiting the present invention.

As shown in Figure 1, the 1/8 pixel precision interpolation method according to the embodiment of the present invention includes the following steps:

S101: Set the sub-pixel category and the position of the sub-pixel;

Sub-pixels include 3 categories, including half-pixels, 1/4 pixels and 1/8 pixels. Half pixel, 1/4 pixel, and 1/8 pixel are subdivided into categories respectively, and can be divided into 14 subcategories in total.

Specifically, as shown in FIG. 3 , the half-pixels include one type of half-pixels and two types of half-pixels.

(1) A class of half-pixels are half-pixels in the same row or column as integer pixels, such as a5 and e1;

(2) The second type of half pixel point is the remaining half pixel point in the half pixel point, that is, other half pixel points except the first type of half pixel point, such as e5;

The 1/4 pixel points include a type 1/4 pixel point, a type 2 1/4 pixel point and a type 3 1/4 pixel point.

(3) Class 1/4 pixels are 1/4 pixels in the same row or column as integer pixels, such as a3, a7, c1, g1;

(4) The second type of 1/4 pixel point is a 1/4 pixel point that is in the same row as the first type of half pixel point and the second type of half pixel point or in the same column as the first type of half pixel point and the second type of half pixel point , such as c5, e3, e7, g5;

(5) The three types of 1/4 pixels are the remaining 1/4 pixels in the 1/4 pixels, that is, the 1/4 pixels except the first type of 1/4 pixels and the second type of 1/4 pixels , such as c3, c7, g3, g7;

1/8 pixels include Class I 1/8 pixels, Class II 1/8 pixels, Class III 1/8 pixels, Class IV 1/8 pixels, Class V 1/8 pixels, Class VI 1/8 8 pixels, seven types of 1/8 pixels, eight types of 1/8 pixels and nine types of 1/8 pixels.

(6) A class 1/8 pixel is a 1/8 pixel located between an integer pixel and a class 1/4 pixel, such as a2, a8, b1, h1;

(7) The second type of 1/8 pixel point is a 1/8 pixel point between the first type of half pixel point and the first type of 1/4 pixel point, such as a4, a6, d1, f1;

(8) Three types of 1/8 pixels are 1/8 pixels between one type of 1/4 pixels and three types of 1/4 pixels, such as b3, b7, c2, c8, g2, g8, h2 , h8;

(9) The four types of 1/8 pixels are 1/8 pixels located between the first type of half pixels and the second type of 1/4 pixels, such as b5, e2, e8, h5;

(10) The five types of 1/8 pixels are 1/8 pixels located between the three types of 1/4 pixels and the second type of 1/4 pixels, such as c4, c6, d3, d7, f3, f7, g4 , g6;

(11) Six types of 1/8 pixels are 1/8 pixels located between the second type of 1/4 pixels and the second type of half pixels, such as d5, e4, e6, f5;

(12) The seven types of 1/8 pixels are 1/8 pixels located between the first type of 1/8 pixels and the third type of 1/8 pixels, such as b2, b8, g2, g8;

(13) Eight types of 1/8 pixels are 1/8 pixels between the second type of 1/8 pixels, the third type of 1/8 pixels, the fourth type of 1/8 pixels, and the fifth type of 1/8 pixels points, such as b4, b6, d2, d8, f2, f8, h4, h6;

(14) The nine types of 1/8 pixels are 1/8 pixels between the fifth types of 1/8 pixels and the sixth types of 1/8 pixels, such as d4, d6, f4, f6.

It can be seen from Figure 3 that the sub-pixels of the same small category have symmetry. For example, d4 and d6 belonging to the nine types of 1/8 pixel points are symmetrical in the vertical direction, and d5 and e4 belonging to the six types of 1/8 pixel points are symmetrical in the diagonal direction.

S102: Acquire integer pixels, and perform interpolation according to the positions of various sub-pixels that have been set to obtain sub-pixels.

Select the integer pixel in the (M+5)×(N+5) integer pixel block closest to the sub-pixel block formed by the sub-pixels to be interpolated from the original reference frame as the reference pixel, and according to the set The positions of various sub-pixels are interpolated to obtain sub-pixels, where M is the number of sub-pixels in each row in the sub-pixel block to be interpolated, and N is the number of sub-pixels in each column of sub-pixels to be interpolated the number of points. In one embodiment of the present invention, M=1, N=1. Thus, for the sub-pixel position above, the 6×6 integer pixels closest to the sub-pixel to be interpolated are used as reference pixels for interpolation. FIG. 2 shows a 6×6 integer pixel block, where A1˜A6, B1˜B6, C1˜C6, D1˜D6, E1˜E6, F1˜F6 are all integer pixels.

Use the interpolation filter to obtain various sub-pixel points. For the interpolation at different sub-pixel positions, the tap coefficient of the interpolation filter part of the filter used is 0, and the corresponding taps can be omitted during implementation, so the actual interpolation filters at sub-pixel positions are all less than 36 taps filter. And the sum of filter tap coefficients corresponding to each type of sub-pixel is 256.

The sub-pixel interpolation acquisition of each sub-category will be specifically described below.

(1) A class of half-pixel points through the tap coefficients are {{0, 0, 0, 0, 0, 0}, {1, -3, 2, 2, -3, 1}, {5, -27, 150 , 150, -27, 5}, {1, -3, 2, 2, -3, 1}, {0, 0, 0, 0, 0, 0}, {0, 0, 0, 0, 0, 0, 0}} 18-tap filter or filter interpolation obtained by performing symmetrical transformation on the filter. in,

a5＝(1*B1+(-3)*B2+2*B3+2*B4+(-3)*B5+2*B6+5*C1+(-27)*C2+150*C3+150*C4+(- 27) *C5+5*C6+1*D1+(-3)*D2+2*D3+2*D4+(-3)*D5+1*D6)/256.

e1＝(1*A2+5*A3+1*A4+(-3)*B2+(-27)*B3+(-3)*B4+2*C2+150*C3+2*C4+2*D2+150 *D3+2*D4+(-3)*E2+(-27)*E3+(-2)*E4+1*F2+5*F3+1*F4)/256.

(2) The tap coefficient of the second type of half-pixel point is

{{0, 0, 3, 3, 0, 0}, {0, 0, -16, -16, 0, 0}, {3, -16, 90, 90, -16, 3}, {3, A 20-tap filter of -16, 90, 90, -16, 3}, {0, 0, -16, -16, 0, 0}, {0, 0, 3, 3, 0, 0}} or The filter interpolation obtained by performing a symmetrical transformation on it. in,

e5＝(3*A3+3*A4+(-16)*B3+(-16)*B4+3*C1+(-16)*C2+90*C3+90*C4+(-16)*C5+3*C6 +3*D1+(-16)*D2+90*D3+90*D4+(-16)*D5+3*D6+(-16)*E3+(-16)*E4+3*F3+3*F4)/ 256.

(3) A class of 1/4 pixel points through the tap coefficient is

{{0, 0, 0, 0, 0, 0}, {1, -2, 1, 1, -2, 1}, {5, -27, 223, 67, -14, 2}, {1, -2, 1, 1, -2, 1}, {0, 0, 0, 0, 0, 0}, {0, 0, 0, 0, 0, 0}} 18-tap filter or its Filter interpolation obtained by performing symmetric transformation. in,

a3＝(1*B1+(-2)*B2+1*B3+1*B4+(-2)*B5+1*B6+5*C1+(-27)*C2+223*C3+67*C4+(- 14) *C5+2*C6+1*D1+(-2)*D2+1*D3+1*D4+(-2)*D5+1*D6)/256.

a7＝(1*B1+(-2)*B2+1*B3+1*B4+(-2)*B5+1*B6+2*C1+(-14)*C2+67*C3+223*C4+(- 27) *C5+5*C6+1*D1+(-2)*D2+1*D3+1*D4+-2*D5+1*D6)/256.

c1＝(1*A2+5*A3+1*A4+(-2)*B2+(-27)*B3+(-2)*B3+1*C2+223*C3+1*C4+1*D2+67 *D3+1*D4+(-2)*E2+(-14)*E3+(-2)*E4+1*F2+2*F3+1*F4)/256.

g1＝(1*A2+2*A3+1*A4+(-2)*B2+(-14)*B3+(-2)*B4+1*C2+67*C3+1*C4+1*D2+223 *D3+1*D4+(-2)*E2+(-27)*E3+(-2)*E4+1*F2+5*F3+1*F4)/256.

(4) The second type of 1/4 pixels through the tap coefficients are {{0, 0, 3, 3, 0, 0}, {0, 0, -15, -15, 0, 0}, {4, -23 , 132, 132, -23, 4}, {2, -9, 41, 41, -9, 2} {0, 0, -9, -9, 0, 0}, {0, 0, 2, 2 , 0, 0}}'s 20-tap filter or filter interpolation obtained by performing symmetrical transformation on it. in,

c5＝(3*A3+3*A4+(-15)*B3+(-15)*B4+4*C1+(-23)*C2+132*C3+132*C4+(-23)*C5+4*C6 +2*D1+(-9)*D2+41*D3+41*D4+(-9)*D5+2*D6+(-9)*E3+(-9)*E4+2*F3+2*F4)/ 256.

e3＝(4*A3+2*A4+(-23)*B3+(-9)*B4+3*C1+(-15)*C2+132*C3+41*C4+(-9)*C5+2*C6 +3*D1+(-15)*D2+132*D3+41*D4+(-9)*D5+2*D6+(-23)*E3+(-9)*E4+4*F3+2*F4)/ 256.

(5) The three types of 1/4 pixel points pass the tap coefficient as

{{0, 0, 4, 2, 0, 0}, {0, 0, -23, -8, 0, 0}, {4, -23, 194, 60, -12, 2}, {2, 20-tap filters of -8, 60, 20, -5, 1}, {0, 0, -12, -5, 0, 0}, {0, 0, 2, 1, 0, 0}} or Filter interpolation obtained by performing symmetrical transformation on the filter. in,

c3＝(4*A3+2*A4+(-23)*B3+(-8)*B4+4*C1+(-23)*C2+194*C3+60*C4+(-12)*C5+2*C6 +2*D1+(-8)*D2+60*D3+20*D4+(05)*D5+1*D6+(-12)*E3+(-5)*E4+2*F3+1*F4)/256 .

c7＝(2*A3+4*A4+(-8)*B3+(-23)*B4+2*C1+(-12)*C2+60*C3+194*C4+(-23)*C5+4*C6 +1*D1+(-5)*D2+20*D3+60*D4+(-8)*D5+2*D6+(-5)*E3+(-12)*E4+1*F3+2*F4)/ 256.

g3＝(2*A3+1*A4+(-12)*B3+(-5)*B4+2*C1+(-8)*C2+60*C3+20*C4+(-5)*C5+1*C6 +4*D1+(-23)*D2+194*D3+60*D4+(-12)*D5+2*D6+(-23)*E3+(-8)*E4+4*F3+2*F4)/ 256.

g7＝(1*A3+2*A4+(-5)*B3+(-12)*B4+1*C1+(-5)*C2+20*C3+60*C4+(-8)*C5+2*C6 +2*D1+(-12)*D2+60*D3+194*D4+(-23)*D5+4*D6+(-8)*E3+(-23)*E4+2*F3+4*F4)/ 256.

(6) A class of 1/8 pixels through the tap coefficient is

{{0, 0, 0, 0, 0, 0}, {0, -1, 1, 1, -1, 0}, {3, -18, 246, 30, -6, 1}, {0, A 14-tap filter of -1, 1, 1, -1, 0}, {0, 0, 0, 0, 0, 0}, {0, 0, 0, 0, 0, 0}} or the Filter interpolation obtained by performing symmetrical transformation on the filter. Among them, a2=((-1)*B2+1*B3+1*B4+(-1)*B5+3*C1+(-18)*C2+246*C3+30*C4+(-6)*C5+ 1*C6+(-1)*D2+1*D3+1*D4+(-1)*D5)/256.

a8＝((-1)*B2+1*B3+1*B4+(-1)*B5+1*C1+(-6)*C2+30*C3+246*C4+(-18)*C5+3* C6+(-1)*D2+1*D3+1*D4+(-1)*D5)/256.

b1＝(3*A3+(-1)*B2+(-18)*B3+(-1)*B4+1*C2+246*C3+1*C4+1*D2+30*D3+1*D4+(- 1)*E2+(-6)*E3+(-1)*E4+1*F3)/256.

h1＝(1*A3+(-1)*B2+(-6)*B3+(-1)*B4+1*C2+30*C3+1*C4+1*D2+246*D3+1*D4+(- 1)*E2+(-18)*E3+(-1)*E4+3*F3)/256.

(7) The second type of 1/8 pixels through the tap coefficient is

{{0, 0, 0, 0, 0, 0}, {1, -3, 2, 1, -2, 1}, {5, -29, 189, 108, -21, 4}, {1, An 18-tap filter of -3, 2, 1, -2, 1}, {0, 0, 0, 0, 0, 0}, {0, 0, 0, 0, 0, 0}} or the Filter interpolation obtained by performing symmetrical transformation on the filter. in,

a4＝(1*B1+(-3)*B2+2*B3+1*B4+(-2)*B5+1*B6+5*C1+(-29)*C2+189*C3+108*C4+(- 21)*C5+4*C6+1*D1+(-3)*D2+2*D3+1*D4+(-2)*D5+1*D6)/256.

a6＝(1*B1+(-2)*B2+1*B3+2*B4+(-3)*B5+1*B6+4*C1+(-21)*C2+108*C3+189*C4+(- 29) *C5+5*C6+1*D1+(-2)*D2+1*D3+2*D4+(-3)*D5+1*D6)/256.

d1＝(1*A2+5*A3+1*A4+(-3)*B2+(-29)*B3+(-3)*B4+2*C2+189*C3+2*C4+1*D2+108 *D3+1*D4+(-2)*E2+(-21)*E3+(-2)*E4+1*F2+4*F3+1*F4)/256.

f1＝(1*A2+4*A3+1*A4+(-2)*B2+(-21)*B3+(-2)*B4+1*C2+108*C3+1*C4+2*D2+189 *D3+2*D4+(-3)*E2+(-29)*E3+(-3)*E4+1*F2+5*F3+1*F4)/256.

(8) The three types of 1/8 pixels pass through the tap coefficient as

{{0, 0, 3, 1, 0, 0}, {1, -1, -14, -4, -1, 0}, {4, -25, 214, 65, -13, 2}, { 1, -5, 27, 9, -3, 1}, {0, 0, -6, -2, 0, 0} {0, 0, 1, 1, 0, 0}} 22-tap filter Or obtain filter interpolation obtained by performing symmetrical transformation on the filter. in,

b3＝(3*A3+1*A4+1*B1+(-1)*B2+(-14)*B3+(-4)*B4+(-1)*B5+4*C1+(-25)*C2+214 *C3+65*C4+(-13)*C5+2*C6+1*D1+(-5)*D2+27*D3+9*D4+(-3)*D5+1*D6+(-6)*E3+ (-2)*E4+1*F3+1*F4)/256.

b7＝(1*A3+3*A4+(-1)*B2+(-4)*B3+(-14)*B4+(-1)*B5+1*B6+2*C1+(-13)*C2+65 *C3+214*C4+(-25)*C5+4*C6+1*D1+(-3)*D2+9*D3+27*D4+(-5)*D5+1*D6+(-2)*E3+ (-6)*E4+1*F3+1*F4)/256.

c2＝(1*A2+4*A3+1*A4+(-1)*B2+(-25)*B3+(-5)*B4+3*C1+(-14)*C2+214*C3+27*C4+ (-6)*C5+1*C6+1*D1+(-4)*D2+65*D3+9*D4+(-2)*D5+1*D6+(-1)*E2+(-13)*E3+ (-3)*E4+2*F3+1*F4)/256.

c8＝(1*A3+4*A4+1*A5+(-5)*B3+(-25)*B4+(-1)*B5+1*C1+(-6)*C2+27*C3+214*C4+ (-14)*C5+3*C6+1*D1+(-2)*D2+9*D3+65*D4+(-4)*D5+1*D6+(-3)*E3+(-13)*E4+ (-1)*E5+1*F3+2*F4)/256.

g2＝(2*A3+1*A4+(-1)*B2+(-13)*B3+(-3)*B4+1*C1+(-4)*C2+65*C3+9*C4+(-2) *C5+1*C6+3*D1+(-14)*D2+214*D3+27*D4+(-6)*D5+1*D6+(-1)*E2+(-25)*E3+(-5) *E4+1*F2+4*F3+1*F4)/256.

g8＝(1*A3+2*A4+(-3)*B3+(-13)*B4+(-1)*B5+1*C1+(-2)*C2+9*C3+65*C4+(-4) *C5+1*C6+1*D1+(-6)*D2+27*D3+214*D4+(-14)*D5+3*D6+(-5)*E3+(-25)*E4+(-1) *E5+1*F3+4*F4+1*F5)/256.

h3＝(1*A3+1*A4+(-6)*B3+(-2)*B4+1*C1+(-5)*C2+27*C3+9*C4+(-3)*C5+1*C6 +4*D1+(-25)*D2+214*D3+65*D4+(-13)*D5+2*D6+1*E1+(-1)*E2+(-14)*E3+(-4)*E4+ (-1)*E5+3*F3+1*F4)/256.

h7＝(1*A3+1*A4+(-2)*B3+(-6)*B4+1*C1+(-3)*C2+9*C3+27*C4+(-5)*C5+1*C6 +2*D1+(-13)*D2+65*D3+214*D4+(-25)*D5+4*D6+(-1)*E2+(-4)*E3+(-14)*E4+(-1) *E5+1*E6+1*F3+3*F4)/256.

(9) The four types of 1/8 pixels pass through the tap coefficients as

{{0, 0, 2, 2, 0, 0}, {1, -1, -9, -9, -1, 1}, {4, -26, 145, 145, -26, 4}, { 1, -5, 19, 19, -5, 1}, {0, 0, -4, -4, 0, 0}, {0, 0, 1, 1, 0, 0}} 24-tap filtering The filter or the filter interpolation obtained by performing a symmetrical transformation on the filter. in,

b5＝(2*A3+2*A4+1*B1+(-1)*B2+(-9)*B3+(-9)*B4+(-1)*B5+1*B6+4*C1+(-26) *C2+145*C3+145*C4+(-26)*C5+4*C6+1*D1+(-5)*D2+19*D3+19*D4+(-5)*D5+1*D6+(- 4)*E3+(-4)*E4+1*F3+1*F4)/256.

e2＝(1*A2+4*A3+1*A4+(-1)*B2+(-26)*B3+(-5)*B4+2*C1+(-9)*C2+145*C3+19*C4+ (-4)*C5+1*C6+2*D1+(-9)*D2+145*D3+19*D4+(-4)*D5+1*D6+(-1)*E2+(-26)*E3+ (-5)*E4+1*F2+4*F3+1*F4)/256.

e8＝(1*A3+4*A4+1*A5+(-5)*B3+(-26)*B4+(-1)*B5+1*C1+(-4)*C2+19*C3+145*C4+ (-9)*C5+2*C6+1*D1+(-4)*D2+19*D3+145*D4+(-9)*D5+2*D6+(-5)*E3+(-26)*E4+ (-1)*E5+1*F3+4*F4+1*F5)/256.

h5＝(1*A3+1*A4+(-4)*B3+(-4)*B4+1*C1+(-5)*C2+19*C3+19*C4+(-5)*C5+1*C6 +4*D1+(-26)*D2+145*D3+145*D4+(-26)*D5+4*D6+1*E1+(-1)*E2+(-9)*E3+(-9)*E4+ (-1)*E5+1*E6+2*F3+2*F4)/256.

(10) The five types of 1/8 pixel points pass the tap coefficient as

{{0, 0, 4, 2, 0, 0}, {0, 0, -19, -11, 0, 0}, {4, -25, 166, 95, -18, 3}, {2, 20-tap filters of -9, 51, 31, -7, 2}, {0, 0, -11, -7, 0, 0}, {0, 0, 2, 1, 0, 0}} or Filter interpolation obtained by performing symmetrical transformation on the filter. in,

c4＝(4*A3+2*A4+(-19)*B3+(-11)*B4+4*C1+(-25)*C2+166*C3+95*C4+(-18)*C5+3*C6 +2*D1+(-9)*D2+51*D3+31*D4+(-7)*D5+2*D6+(-11)*E3+(-7)*E4+2*F3+1*F4)/ 256.

c6＝(2*A3+4*A4+(-11)*B3+(-19)*B4+3*C1+(-18)*C2+95*C3+166*C4+(-25)*C5+4*C6 +2*D1+(-7)*D2+31*D3+51*D4+(-9)*D5+2*D6+(-7)*E3+(-11)*E4+1*F3+2*F4)/ 256.

d3＝(4*A3+2*A4+(-25)*B3+(-9)*B4+4*C1+(-19)*C2+166*C3+51*C4+(-11)*C5+2*C6 +2*D1+(-11)*D2+95*D3+31*D4+(-7)*D5+1*D6+(-18)*E3+(-7)*E4+3*F3+2*F4)/ 256.

d7＝(2*A3+4*A4+(-9)*B3+(-25)*B4+2*C1+(-11)*C2+51*C3+166*C4+(-19)*C5+4*C6 +1*D1+(-7)*D2+31*D3+95*D4+(-11)*D5+2*D6+(-7)*E3+(-18)*E4+2*F3+3*F4)/ 256.

f3＝(3*A3+2*A4+(-18)*B3+(-7)*B4+2*C1+(-11)*C2+95*C3+31*C4+(-7)*C5+1*C6 +4*D1+(-19)*D2+166*D3+51*D4+(-11)*D5+2*D6+(-25)*E3+(-9)*E4+4*F3+2*F4)/ 256.

f7＝(2*A3+3*A4+(-7)*B3+(-18)*B4+1*C1+(-7)*C2+31*C3+95*C4+(-11)*C5+2*C6 +2*D1+(-11)*D2+51*D3+166*D4+(-19)*D5+4*D6+(-9)*E3+(-25)*E4+2*F3+4*F4)/ 256.

g4＝(2*A3+1*A4+(-11)*B3+(-7)*B4+2*C1+(-9)*C2+51*C3+31*C4+(-7)*C5+2*C6 +4*D1+(-25)*D2+166*D3+95*D4+(-18)*D5+3*D6+(-19)*E3+(-11)*E4+4*F3+2*F4)/ 256.

g6＝(1*A3+2*A4+(-7)*B3+(-11)*B4+2*C1+(-7)*C2+31*C3+51*C4+(-9)*C5+2*C6 +3*D1+(-18)*D2+95*D3+166*D4+(-25)*D5+4*D6+(-11)*E3+(-19)*E4+2*F3+4*F4)/ 256.

(11) Six types of 1/8 pixel points pass the tap coefficient as

{{0, 0, 3, 3, 0, 0}, {0, 0, -17, -17, 0, 0}, {4, -20, 113, 113, -20, 4}, {2, 20-tap filters of -12, 66, 66, -12, 2}, {0, 0, -13, -13, 0, 0}, {0, 0, 2, 2, 0, 0}} or Filter interpolation obtained by performing symmetrical transformation on the filter. in,

d5＝(3*A3+3*A4+(-17)*B3+(-17)*B4+4*C1+(-20)*C2+113*C3+113*C4+(-20)*C5+4*C6 +2*D1+(-12)*D2+66*D3+66*D4+(-12)*D5+2*D6+(-13)*E3+(-13)*E4+2*F3+2*F4)/ 256.

e4＝(4*A3+2*A4+(-20)*B3+(-12)*B4+3*C1+(-17)*C2+113*C3+66*C4+(-13)*C5+2*C6 +3*D1+(-17)*D2+113*D3+66*D4+(-13)*D5+2*D6+(-20)*E3+(-12)*E4+4*F3+2*F4)/ 256.

e6＝(2*A3+4*A4+(-12)*B3+(-20)*B4+2*C1+(-13)*C2+66*C3+113*C4+(-17)*C5+3*C6 +2*D1+(-13)*D2+66*D3+113*D4+(-17)*D5+3*D6+(-12)*E3+(-20)*E4+2*F3+4*F4)/ 256.

f5＝(2*A3+2*A4+(-13)*B3+(-13)*B4+2*C1+(-12)*C2+66*C3+66*C4+(-12)*C5+2*C6 +4*D1+(-20)*D2+113*D3+113*D4+(-20)*D5+4*D6+(-17)*E3+(-17)*E4+3*F3+3*F4)/ 256.

(12) Seven types of 1/8 pixel points pass the tap coefficient as

{{0, 0, 3, 1, 0, 0}, {0, -1, -17, -2, 0, 0}, {3, -17, 237, 29, -6, 1}, {1 , -2, 29, 5, -2, 0}, {0, 0, -6, -1, 0, 0}, {0, 0, 1, 0, 0, 0}} 19-tap filter Or obtain filter interpolation obtained by performing symmetrical transformation on the filter. in,

b2＝(3*A3+1*A4+(-1)*B2+(-17)*B3+(-2)*B4+3*C1+(-17)*C2+237*C3+29*C4+(-6) *C5+1*C6+1*D1+(-2)*D2+29*D3+5*D4+(-2)*D5+(-6)*E3+(-1)*E4+1*F3)/256.

b8＝(1*A3+3*A4+(-2)*B3+(-17)*B4+(-1)*B5+1*C1+(-6)*C2+29*C3+237*C4+(-17) *C5+3*C6+(-1)*D2+5*D3+29*D4+(-2)*D5+1*D6+(-2)*E3+(-6)*E4+1*F4)/256.

h2＝(1*A3+(-6)*B3+(-1)*B4+1*C1+(-2)*C2+29*C3+5*C4+(-2)*C5+3*D1+(-17) *D2+237*D3+29*D4+(-6)*D5+1*D6+(-1)*E2+(-17)*E3+(-2)*E4+3*F3+1*F4)/256.

h8＝(1*A4+(-2)*B3+(-6)*B4+(-1)*C2+5*C3+29*C4+(-2)*C5+1*C6+1*D1+(-6) *D2+29*D3+237*D4+(-17)*D5+3*D6+(-2)*E3+(-17)*E4+(-1)*E5+1*F3+3*F4)/256.

(13) The eight types of 1/8 pixel points pass the tap coefficient as

{{0, 0, 2, 1, 0, 0}, {1, -1, -12, -7, -1, 0}, {5, -28, 182, 105, -20, 4}, { 1, -5, 24, 15, -5, 1}, {0, 0, -5, -3, 0, 0}, {0, 0, 1, 1, 0, 0}} 23-tap filtering The filter or the filter interpolation obtained by performing a symmetrical transformation on the filter. in,

b4＝(2*A3+1*A4+1*B1+(-1)*B2+(-12)*B3+(-7)*B4+(-1)*B5+5*C1+(-28)*C2+182 *C3+105*C4+(-20)*C5+4*C6+1*D1+(-5)*D2+24*D3+15*D4+(-5)*D5+1*D6+(-5)*E3+ (-3)*E4+1*F3+1*F4)/256.

b6＝(1*A3+2*A4+(-1)*B2+(-7)*B3+(-12)*B4+(-1)*B5+1*B6+4*C1+(-20)*C2+105 *C3+182*C4+(-28)*C5+5*C6+1*D1+(-5)*D2+15*D3+24*D4+(-5)*D5+1*D6+(-3)*E3+ (-5)*E4+1*F3+1*F4)/256.

d2＝(1*A2+5*A3+1*A4+(-1)*B2+(-28)*B3+(-5)*B4+2*C1+(-12)*C2+182*C3+24*C4+ (-5)*C5+1*C6+1*D1+(-7)*D2+105*D3+15*D4+(-3)*D5+1*D6+(-1)*E2+(-20)*E3+ (-5)*E4+4*F3+1*F4)/256.

d8＝(1*A3+5*A4+1*A5+(-5)*B3+(-28)*B4+(-1)*B5+1*C1+(-5)*C2+24*C3+182*C4+ (-12)*C5+2*C6+1*D1+(-3)*D2+15*D3+105*D4+(-7)*D5+1*D6+(-5)*E3+(-20)*E4+ (-1)*E5+1*F3+4*F4)/256.

f2＝(4*A3+1*A4+(-1)*B2+(-20)*B3+(-5)*B4+1*C1+(-7)*C2+105*C3+15*C4+(-3) *C5+1*C6+2*D1+(-12)*D2+182*D3+24*D4+(-5)*D5+1*D6+(-1)*E2+(-28)*E3+(-5) *E4+1*F2+5*F3+1*F4)/256.

f8＝(1*A3+4*A4+(-5)*B3+(-20)*B4+(-1)*B5+1*C1+(-3)*C2+15*C3+105*C4+(-7) *C5+1*C6+1*D1+(-5)*D2+24*D3+182*D4+(-12)*D5+2*D6+(-5)*E3+(-28)*E4+(-1) *E5+1*F3+5*F4+1*F5)/256.

h4＝(1*A3+1*A4+(-5)*B3+(-3)*B4+1*C1+(-5)*C2+24*C3+15*C4+(-5)*C5+1*C6 +5*D1+(-28)*D2+182*D3+105*D4+(-20)*D5+4*D6+1*E1+(-1)*E2+(-12)*E3+(-7)*E4+ (-1)*E5+2*F3+1*F4)/256.

h6＝(1*A3+1*A4+(-3)*B3+(-5)*B4+1*C1+(-5)*C2+15*C3+24*C4+(-5)*C5+1*C6 +4*D1+(-20)*D2+105*D3+182*D4+(-28)*D5+5*D6+(-1)*E2+(-7)*E3+(-12)*E4+(-1) *E5+1*E6+1*F3+2*F4)/256.

(14) Nine types of 1/8 pixels pass the tap coefficient as

{{0, 0, 4, 3, 0, 0}, {0, 0, -21, -13, 0, 0}, {4, -21, 141, 82, -16, 3}, {2, 20-tap filters of -13, 82, 48, -10, 2}, {0, 0, -16, -10, 0, 0}, {0, 0, 3, 2, 0, 0}} or Filter interpolation obtained by performing symmetrical transformation on the filter. in,

d4＝(4*A3+3*A4+(-21)*B3+(-13)*B4+4*C1+(-21)*C2+141*C3+82*C4+(-16)*C5+3*C6 +2*D1+(-13)*D2+82*D3+48*D4+(-10)*D5+2*D6+(-16)*E3+(-10)*E4+3*F3+2*F4)/ 256.

d6＝(3*A3+4*A4+(-13)*B3+(-21)*B4+3*C1+(-16)*C2+82*C3+141*C4+(-21)*C5+4*C6 +2*D1+(-10)*D2+48*D3+82*D4+(-13)*D5+2*D6+(-10)*E3+(-16)*E4+2*F3+3*F4)/ 256.

f4＝(3*A3+2*A4+(-16)*B3+(-10)*B4+2*C1+(-13)*C2+82*C3+48*C4+(-10)*C5+2*C6 +4*D1+(-21)*D2+141*D3+82*D4+(-16)*D5+3*D6+(-21)*E3+(-13)*E4+4*F3+3*F4)/ 256.

f6＝(2*A3+3*A4+(-10)*B3+(-16)*B4+2*C1+(-10)*C2+48*C3+82*C4+(-13)*C5+3*C6 +3*D1+(-16)*D2+82*D3+141*D4+(-21)*D5+4*D6+(-13)*E3+(-21)*E4+2*F3+4*F4)/ 256.

According to the interpolation method with 1/8 pixel precision in the embodiment of the present invention, for all sub-pixel position interpolation, only integer pixels need to be used as reference pixels, which simplifies the interpolation process in the sub-pixel position, thereby reducing the time delay of the interpolation process. At the same time, a high-precision interpolation filter is used, which can make the motion vector accurate to 1/8 pixel level in motion compensation, improve the prediction accuracy, reduce the bit rate, and improve the compression efficiency of video coding.

The embodiment of the present invention described below with reference to FIG. 4 also proposes a 1/8 pixel precision interpolation device 100 .

As shown in FIG. 4 , the 1/8 pixel precision interpolation device 100 according to the embodiment of the present invention includes an acquisition module 110 , a sub-pixel interpolation module 120 and an output module 130 . Wherein, the acquisition module 110 is connected to the sub-pixel interpolation module 120 through a data bus, and the sub-pixel interpolation module 120 is connected to the output module 130 through a data bus.

The acquisition module 110 selects an integer pixel point in the (M+5)×(N+5) integer pixel block closest to the sub-pixel block composed of sub-pixel points to be obtained by interpolation from the original reference frame as a reference pixel point. Wherein, M is the number of sub-pixels in each row in the sub-pixel block to be obtained by interpolation, and N is the number of sub-pixels in each column of the sub-pixels to be obtained by interpolation. In one embodiment of the present invention, M=1, N=1. Thus, for the sub-pixel position above, the 6×6 integer pixels closest to the sub-pixel to be interpolated are used as reference pixels for interpolation. FIG. 2 shows a 6×6 integer pixel block, where A1˜A6, B1˜B6, C1˜C6, D1˜D6, E1˜E6, F1˜F6 are all integer pixels.

The sub-pixel interpolation module 120 is connected to the acquisition module 110, and performs interpolation on the integer pixel points from the acquisition module 110 as reference pixels to obtain sub-pixel points. Sub-pixels include 3 categories, including half-pixels, 1/4 pixels and 1/8 pixels. Half pixel, 1/4 pixel, and 1/8 pixel are subdivided into categories respectively, and can be divided into 14 subcategories in total. In order to obtain sub-pixel points corresponding to each subclass, as shown in Figure 5, the sub-pixel interpolation module 120 further includes a first-class half-pixel interpolation unit 1201, a second-class half-pixel interpolation unit 1202, a first-class 1/4 pixel interpolation unit 1203, the second type 1/4 pixel interpolation unit 1204, the third type 1/4 pixel interpolation unit 1205, the first type 1/8 pixel interpolation unit 1206, the second type 1/8 pixel interpolation unit 1207, the third type 1/8 pixel interpolation unit 1208, four types of 1/8 pixel interpolation unit 1209, five types of 1/8 pixel interpolation unit 1210, six types of 1/8 pixel interpolation unit 1211, seven types of 1/8 pixel interpolation unit 1212, eight types of 1/8 pixel interpolation unit 1213. A nine-type 1/8 pixel interpolation unit 1214. The acquisition module 110 and the output module 130 are respectively connected to each interpolation unit through a data bus. In an embodiment of the present invention, each interpolation unit mentioned above may be a filter, the sum of tap coefficients of each filter is 256, and the number of taps of each filter is less than 36.

Each interpolation unit will be described below with reference to FIG. 5 .

(1) A class of half-pixel interpolation unit 1201 can interpolate to obtain a class of half-pixels, which are half-pixels in the same row or column as integer pixels, such as a5 and e1. The filter of a type of half pixel interpolation unit 1201 has tap coefficients {{0, 0, 0, 0, 0, 0}, {1, -3, 2, 2, -3, 1}, {5, -27 , 150, 150, -27, 5}, {1, -3, 2, 2, -3, 1}, {0, 0, 0, 0, 0, 0}, {0, 0, 0, 0, 0, 0, 0}} an 18-tap filter. in,

a5＝(1*B1+(-3)*B2+2*B3+2*B4+(-3)*B5+2*B6+5*C1+(-27)*C2+150*C3+150*C4+(- 27) *C5+5*C6+1*D1+(-3)*D2+2*D3+2*D4+(-3)*D5+1*D6)/256.

e1＝(1*A2+5*A3+1*A4+(-3)*B2+(-27)*B3+(-3)*B4+2*C2+150*C3+2*C4+2*D2+150 *D3+2*D4+(-3)*E2+(-27)*E3+(-2)*E4+1*F2+5*F3+1*F4)/256.

(2) The second-type half-pixel interpolation unit 1202 can interpolate to obtain the second-type half-pixels, the second-type half-pixels are the remaining half-pixels in the half-pixels, that is, other half-pixels except the first-type half-pixels, such as e5. The filter of the second type of half-pixel interpolation unit 1202 has a tap coefficient of

{{0, 0, 3, 3, 0, 0}, {0, 0, -16, -16, 0, 0}, {3, -16, 90, 90, -16, 3}, {3, -16, 90, 90, -16, 3}, {0, 0, -16, -16, 0, 0}, {0, 0, 3, 3, 0, 0}} 20-tap filter. in,

e5＝(3*A3+3*A4+(-16)*B3+(-16)*B4+3*C1+(-16)*C2+90*C3+90*C4+(-16)*C5+3*C6 +3*D1+(-16)*D2+90*D3+90*D4+(-16)*D5+3*D6+(-16)*E3+(-16)*E4+3*F3+3*F4)/ 256.

(3) A type of 1/4 pixel interpolation unit 1203 can interpolate to obtain a type of 1/4 pixel point, a type of 1/4 pixel point is a 1/4 pixel point in the same row or column as an integer pixel point, such as a3 , a7, c1, g1. The filter of a kind of 1/4 pixel interpolation unit 1203 is that the tap coefficient is

{{0, 0, 0, 0, 0, 0}, {1, -2, 1, 1, -2, 1}, {5, -27, 223, 67, -14, 2}, {1, -2, 1, 1, -2, 1}, {0, 0, 0, 0, 0, 0}, {0, 0, 0, 0, 0, 0}} 20-tap filter. in,

a3＝(1*B1+(-2)*B2+1*B3+1*B4+(-2)*B5+1*B6+5*C1+(-27)*C2+223*C3+67*C4+(- 14) *C5+2*C6+1*D1+(-2)*D2+1*D3+1*D4+(-2)*D5+1*D6)/256.

a7＝(1*B1+(-2)*B2+1*B3+1*B4+(-2)*B5+1*B6+2*C1+(-14)*C2+67*C3+223*C4+(- 27) *C5+5*C6+1*D1+(-2)*D2+1*D3+1*D4+-2*D5+1*D6)/256.

c1＝(1*A2+5*A3+1*A4+(-2)*B2+(-27)*B3+(-2)*B3+1*C2+223*C3+1*C4+1*D2+67 *D3+1*D4+(-2)*E2+(-14)*E3+(-2)*E4+1*F2+2*F3+1*F4)/256.

g1＝(1*A2+2*A3+1*A4+(-2)*B2+(-14)*B3+(-2)*B4+1*C2+67*C3+1*C4+1*D2+223 *D3+1*D4+(-2)*E2+(-27)*E3+(-2)*E4+1*F2+5*F3+1*F4)/256.

(4) The second type of 1/4 pixel interpolation unit 1204 can interpolate to obtain the second type of 1/4 pixel point, the second type of 1/4 pixel point is in the same row with the first type of half pixel point and the second type of half pixel point at the same time or simultaneously with the second type of half pixel point The 1/4 pixels in the same column of the first half pixel and the second half pixel, such as c5, e3, e7, g5. The filter of the second type 1/4 pixel interpolation unit 1204 has tap coefficients {{0, 0, 3, 3, 0, 0}, {0, 0, -15, -15, 0, 0}, {4, -23, 132, 132, -23, 4}, {2, -9, 41, 41, -9, 2} {0, 0, -9, -9, 0, 0}, {0, 0, 2 , 2, 0, 0}} is a 20-tap filter. in,

c5＝(3*A3+3*A4+(-15)*B3+(-15)*B4+4*C1+(-23)*C2+132*C3+132*C4+(-23)*C5+4*C6 +2*D1+(-9)*D2+41*D3+41*D4+(-9)*D5+2*D6+(-9)*E3+(-9)*E4+2*F3+2*F4)/ 256.

e3＝(4*A3+2*A4+(-23)*B3+(-9)*B4+3*C1+(-15)*C2+132*C3+41*C4+(-9)*C5+2*C6 +3*D1+(-15)*D2+132*D3+41*D4+(-9)*D5+2*D6+(-23)*E3+(-9)*E4+4*F3+2*F4)/ 256.

(5) The three types of 1/4 pixel interpolation unit 1205 can interpolate to obtain three types of 1/4 pixels, and the three types of 1/4 pixels are the remaining 1/4 pixels in the 1/4 pixels, that is, except for one type 1 /4 pixels and 1/4 pixels other than the second type of 1/4 pixels, such as c3, c7, g3, g7. The filter of the three types of 1/4 pixel interpolation unit 1205 has a tap coefficient of

{{0, 0, 4, 2, 0, 0}, {0, 0, -23, -8, 0, 0}, {4, -23, 194, 60, -12, 2}, {2, 20-tap filter of -8, 60, 20, -5, 1}, {0, 0, -12, -5, 0, 0}, {0, 0, 2, 1, 0, 0}}. in,

c3＝(4*A3+2*A4+(-23)*B3+(-8)*B4+4*C1+(-23)*C2+194*C3+60*C4+(-12)*C5+2*C6 +2*D1+(-8)*D2+60*D3+20*D4+(05)*D5+1*D6+(-12)*E3+(-5)*E4+2*F3+1*F4)/256 .

c7＝(2*A3+4*A4+(-8)*B3+(-23)*B4+2*C 1+(-12)*C2+60*C3+194*C4+(-23)*C5+4 *C6+1*D1+(-5)*D2+20*D3+60*D4+(-8)*D5+2*D6+(-5)*E3+(-12)*E4+1*F3+2*F4 )/256.

g3＝(2*A3+1*A4+(-12)*B3+(-5)*B4+2*C1+(-8)*C2+60*C3+20*C4+(-5)*C5+1*C6 +4*D1+(-23)*D2+194*D3+60*D4+(-12)*D5+2*D6+(-23)*E3+(-8)*E4+4*F3+2*F4)/ 256.

g7＝(1*A3+2*A4+(-5)*B3+(-12)*B4+1*C1+(-5)*C2+20*C3+60*C4+(-8)*C5+2*C6 +2*D1+(-12)*D2+60*D3+194*D4+(-23)*D5+4*D6+(-8)*E3+(-23)*E4+2*F3+4*F4)/ 256.

(6) The first-class 1/8 pixel interpolation unit 1206 can interpolate to obtain a first-class 1/8 pixel point, a first-class 1/8 pixel point is a 1/8 pixel point between an integer pixel and a first-class 1/4 pixel point , such as a2, a8, b1, h1. The filter of a kind of 1/8 pixel interpolation unit 1206 is that the tap coefficient is

{{0, 0, 0, 0, 0, 0}, {0, -1, 1, 1, -1, 0}, {3, -18, 246, 30, -6, 1}, {0, -1, 1, 1, -1, 0}, {0, 0, 0, 0, 0, 0}, {0, 0, 0, 0, 0, 0}} 14-tap filter. in,

a2＝((-1)*B2+1*B3+1*B4+(-1)*B5+3*C1+(-18)*C2+246*C3+30*C4+(-6)*C5+1* C6+(-1)*D2+1*D3+1*D4+(-1)*D5)/256.

a8＝((-1)*B2+1*B3+1*B4+(-1)*B5+1*C1+(-6)*C2+30*C3+246*C4+(-18)*C5+3* C6+(-1)*D2+1*D3+1*D4+(-1)*D5)/256.

b1＝(3*A3+(-1)*B2+(-18)*B3+(-1)*B4+1*C2+246*C3+1*C4+1*D2+30*D3+1*D4+(- 1)*E2+(-6)*E3+(-1)*E4+1*F3)/256.

h1＝(1*A3+(-1)*B2+(-6)*B3+(-1)*B4+1*C2+30*C3+1*C4+1*D2+246*D3+1*D4+(- 1)*E2+(-18)*E3+(-1)*E4+3*F3)/256.

(7) The second-type 1/8 pixel interpolation unit 1207 can interpolate to obtain the second-type 1/8 pixel point, and the second-type 1/8 pixel point is 1/8 between the first-type half-pixel point and the first-type 1/4 pixel point 8 pixels, such as a4, a6, d1, f1. The filter of the second class 1/8 pixel interpolation unit 1207 is that the tap coefficient is

{{0, 0, 0, 0, 0, 0}, {1, -3, 2, 1, -2, 1}, {5, -29, 189, 108, -21, 4}, {1, -3, 2, 1, -2, 1}, {0, 0, 0, 0, 0, 0}, {0, 0, 0, 0, 0, 0}} 18-tap filter. in,

a4＝(1*B1+(-3)*B2+2*B3+1*B4+(-2)*B5+1*B6+5*C1+(-29)*C2+189*C3+108*C4+(- 21)*C5+4*C6+1*D1+(-3)*D2+2*D3+1*D4+(-2)*D5+1*D6)/256.

a6＝(1*B1+(-2)*B2+1*B3+2*B4+(-3)*B5+1*B6+4*C1+(-21)*C2+108*C3+189*C4+(- 29) *C5+5*C6+1*D1+(-2)*D2+1*D3+2*D4+(-3)*D5+1*D6)/256.

d1＝(1*A2+5*A3+1*A4+(-3)*B2+(-29)*B3+(-3)*B4+2*C2+189*C3+2*C4+1*D2+108 *D3+1*D4+(-2)*E2+(-21)*E3+(-2)*E4+1*F2+4*F3+1*F4)/256.

f1＝(1*A2+4*A3+1*A4+(-2)*B2+(-21)*B3+(-2)*B4+1*C2+108*C3+1*C4+2*D2+189 *D3+2*D4+(-3)*E2+(-29)*E3+(-3)*E4+1*F2+5*F3+1*F4)/256.

(8) The three types of 1/8 pixel interpolation unit 1208 can interpolate to obtain three types of 1/8 pixels, and the three types of 1/8 pixels are located between the first type of 1/4 pixels and the third types of 1/4 pixels 1/8 pixel, such as b3, b7, c2, c8, g2, g8, h2, h8. The filter of the three types of 1/8 pixel interpolation unit 1208 has a tap coefficient of

{{0, 0, 3, 1, 0, 0}, {1, -1, -14, -4, -1, 0}, {4, -25, 214, 65, -13, 2}, { 1, -5, 27, 9, -3, 1}, {0, 0, -6, -2, 0, 0} {0, 0, 1, 1, 0, 0}} 22-tap filter . in,

b3＝(3*A3+1*A4+1*B1+(-1)*B2+(-14)*B3+(-4)*B4+(-1)*B5+4*C1+(-25)*C2+214 *C3+65*C4+(-13)*C5+2*C6+1*D1+(-5)*D2+27*D3+9*D4+(-3)*D5+1*D6+(-6)*E3+ (-2)*E4+1*F3+1*F4)/256.

b7＝(1*A3+3*A4+(-1)*B2+(-4)*B3+(-14)*B4+(-1)*B5+1*B6+2*C1+(-13)*C2+65 *C3+214*C4+(-25)*C5+4*C6+1*D1+(-3)*D2+9*D3+27*D4+(-5)*D5+1*D6+(-2)*E3+ (-6)*E4+1*F3+1*F4)/256.

c2＝(1*A2+4*A3+1*A4+(-1)*B2+(-25)*B3+(-5)*B4+3*C1+(-14)*C2+214*C3+27*C4+ (-6)*C5+1*C6+1*D1+(-4)*D2+65*D3+9*D4+(-2)*D5+1*D6+(-1)*E2+(-13)*E3+ (-3)*E4+2*F3+1*F4)/256.

c8＝(1*A3+4*A4+1*A5+(-5)*B3+(-25)*B4+(-1)*B5+1*C1+(-6)*C2+27*C3+214*C4+ (-14)*C5+3*C6+1*D1+(-2)*D2+9*D3+65*D4+(-4)*D5+1*D6+(-3)*E3+(-13)*E4+ (-1)*E5+1*F3+2*F4)/256.

g2＝(2*A3+1*A4+(-1)*B2+(-13)*B3+(-3)*B4+1*C1+(-4)*C2+65*C3+9*C4+(-2) *C5+1*C6+3*D1+(-14)*D2+214*D3+27*D4+(-6)*D5+1*D6+(-1)*E2+(-25)*E3+(-5) *E4+1*F2+4*F3+1*F4)/256.

g8＝(1*A3+2*A4+(-3)*B3+(-13)*B4+(-1)*B5+1*C1+(-2)*C2+9*C3+65*C4+(-4) *C5+1*C6+1*D1+(-6)*D2+27*D3+214*D4+(-14)*D5+3*D6+(-5)*E3+(-25)*E4+(-1) *E5+1*F3+4*F4+1*F5)/256.

h3＝(1*A3+1*A4+(-6)*B3+(-2)*B4+1*C1+(-5)*C2+27*C3+9*C4+(-3)*C5+1*C6 +4*D1+(-25)*D2+214*D3+65*D4+(-13)*D5+2*D6+1*E1+(-1)*E2+(-14)*E3+(-4)*E4+ (-1)*E5+3*F3+1*F4)/256.

h7＝(1*A3+1*A4+(-2)*B3+(-6)*B4+1*C1+(-3)*C2+9*C3+27*C4+(-5)*C5+1*C6 +2*D1+(-13)*D2+65*D3+214*D4+(-25)*D5+4*D6+(-1)*E2+(-4)*E3+(-14)*E4+(-1) *E5+1*E6+1*F3+3*F4)/256.

(9) The four types of 1/8 pixel interpolation unit 1209 can interpolate to obtain four types of 1/8 pixels, and the four types of 1/8 pixels are 1/8 pixels located between the first type of half pixel and the second type of 1/4 pixel 8 pixels, such as b5, e2, e8, h5. The filter of the four types of 1/8 pixel interpolation unit 1209 has a tap coefficient of

{{0, 0, 2, 2, 0, 0}, {1, -1, -9, -9, -1, 1}, {4, -26, 145, 145, -26, 4}, { 1, -5, 19, 19, -5, 1}, {0, 0, -4, -4, 0, 0}, {0, 0, 1, 1, 0, 0}} 24-tap filtering device. in,

b5＝(2*A3+2*A4+1*B1+(-1)*B2+(-9)*B3+(-9)*B4+(-1)*B5+1*B6+4*C1+(-26) *C2+145*C3+145*C4+(-26)*C5+4*C6+1*D1+(-5)*D2+19*D3+19*D4+(-5)*D5+1*D6+(- 4)*E3+(-4)*E4+1*F3+1*F4)/256.

e2＝(1*A2+4*A3+1*A4+(-1)*B2+(-26)*B3+(-5)*B4+2*C1+(-9)*C2+145*C3+19*C4+ (-4)*C5+1*C6+2*D1+(-9)*D2+145*D3+19*D4+(-4)*D5+1*D6+(-1)*E2+(-26)*E3+ (-5)*E4+1*F2+4*F3+1*F4)/256.

e8＝(1*A3+4*A4+1*A5+(-5)*B3+(-26)*B4+(-1)*B5+1*C1+(-4)*C2+19*C3+145*C4+ (-9)*C5+2*C6+1*D1+(-4)*D2+19*D3+145*D4+(-9)*D5+2*D6+(-5)*E3+(-26)*E4+ (-1)*E5+1*F3+4*F4+1*F5)/256.

h5＝(1*A3+1*A4+(-4)*B3+(-4)*B4+1*C1+(-5)*C2+19*C3+19*C4+(-5)*C5+1*C6 +4*D1+(-26)*D2+145*D3+145*D4+(-26)*D5+4*D6+1*E1+(-1)*E2+(-9)*E3+(-9)*E4+ (-1)*E5+1*E6+2*F3+2*F4)/256.

(10) The five types of 1/8 pixel interpolation unit 1210 can interpolate to obtain five types of 1/8 pixels, and the five types of 1/8 pixels are located between the three types of 1/4 pixels and the second type of 1/4 pixels 1/8 pixel, such as c4, c6, d3, d7, f3, f7, g4, g6. The filter of the five-type 1/8 pixel interpolation unit 1210 is a tap coefficient of

{{0, 0, 4, 2, 0, 0}, {0, 0, -19, -11, 0, 0}, {4, -25, 166, 95, -18, 3}, {2, -9, 51, 31, -7, 2}, {0, 0, -11, -7, 0, 0}, {0, 0, 2, 1, 0, 0}} 20-tap filter. in,

c4＝(4*A3+2*A4+(-19)*B3+(-11)*B4+4*C1+(-25)*C2+166*C3+95*C4+(-18)*C5+3*C6 +2*D1+(-9)*D2+51*D3+31*D4+(-7)*D5+2*D6+(-11)*E3+(-7)*E4+2*F3+1*F4)/ 256.

c6＝(2*A3+4*A4+(-11)*B3+(-19)*B4+3*C1+(-18)*C2+95*C3+166*C4+(-25)*C5+4*C6 +2*D1+(-7)*D2+31*D3+51*D4+(-9)*D5+2*D6+(-7)*E3+(-11)*E4+1*F3+2*F4)/ 256.

d3＝(4*A3+2*A4+(-25)*B3+(-9)*B4+4*C1+(-19)*C2+166*C3+51*C4+(-11)*C5+2*C6 +2*D1+(-11)*D2+95*D3+31*D4+(-7)*D5+1*D6+(-18)*E3+(-7)*E4+3*F3+2*F4)/ 256.

d7＝(2*A3+4*A4+(-9)*B3+(-25)*B4+2*C1+(-11)*C2+51*C3+166*C4+(-19)*C5+4*C6 +1*D1+(-7)*D2+31*D3+95*D4+(-11)*D5+2*D6+(-7)*E3+(-18)*E4+2*F3+3*F4)/ 256.

f3＝(3*A3+2*A4+(-18)*B3+(-7)*B4+2*C1+(-11)*C2+95*C3+31*C4+(-7)*C5+1*C6 +4*D1+(-19)*D2+166*D3+51*D4+(-11)*D5+2*D6+(-25)*E3+(-9)*E4+4*F3+2*F4)/ 256.

f7＝(2*A3+3*A4+(-7)*B3+(-18)*B4+1*C1+(-7)*C2+31*C3+95*C4+(-11)*C5+2*C6 +2*D1+(-11)*D2+51*D3+166*D4+(-19)*D5+4*D6+(-9)*E3+(-25)*E4+2*F3+4*F4)/ 256.

g4＝(2*A3+1*A4+(-11)*B3+(-7)*B4+2*C1+(-9)*C2+51*C3+31*C4+(-7)*C5+2*C6 +4*D1+(-25)*D2+166*D3+95*D4+(-18)*D5+3*D6+(-19)*E3+(-11)*E4+4*F3+2*F4)/ 256.

g6＝(1*A3+2*A4+(-7)*B3+(-11)*B4+2*C1+(-7)*C2+31*C3+51*C4+(-9)*C5+2*C6 +3*D1+(-18)*D2+95*D3+166*D4+(-25)*D5+4*D6+(-11)*E3+(-19)*E4+2*F3+4*F4)/ 256.

(11) The six types of 1/8 pixel interpolation unit 1211 can interpolate to obtain six types of 1/8 pixels, and the six types of 1/8 pixels are 1/8 pixels between the second type of 1/4 pixels and the second type of half pixels 8 pixels, such as d5, e4, e6, f5. The filter of the six types of 1/8 pixel interpolation unit 1211 is a tap coefficient of

{{0, 0, 3, 3, 0, 0}, {0, 0, -17, -17, 0, 0}, {4, -20, 113, 113, -20, 4}, {2, -12, 66, 66, -12, 2}, {0, 0, -13, -13, 0, 0}, {0, 0, 2, 2, 0, 0}} 20-tap filter. in,

d5＝(3*A3+3*A4+(-17)*B3+(-17)*B4+4*C1+(-20)*C2+113*C3+113*C4+(-20)*C5+4*C6 +2*D1+(-12)*D2+66*D3+66*D4+(-12)*D5+2*D6+(-13)*E3+(-13)*E4+2*F3+2*F4)/ 256.

e4＝(4*A3+2*A4+(-20)*B3+(-12)*B4+3*C1+(-17)*C2+113*C3+66*C4+(-13)*C5+2*C6 +3*D1+(-17)*D2+113*D3+66*D4+(-13)*D5+2*D6+(-20)*E3+(-12)*E4+4*F3+2*F4)/ 256.

e6＝(2*A3+4*A4+(-12)*B3+(-20)*B4+2*C1+(-13)*C2+66*C3+113*C4+(-17)*C5+3*C6 +2*D1+(-13)*D2+66*D3+113*D4+(-17)*D5+3*D6+(-12)*E3+(-20)*E4+2*F3+4*F4)/ 256.

f5＝(2*A3+2*A4+(-13)*B3+(-13)*B4+2*C1+(-12)*C2+66*C3+66*C4+(-12)*C5+2*C6 +4*D1+(-20)*D2+113*D3+113*D4+(-20)*D5+4*D6+(-17)*E3+(-17)*E4+3*F3+3*F4)/ 256.

(12) The seven types of 1/8 pixel interpolation unit 1212 can interpolate to obtain seven types of 1/8 pixel points, and the seven types of 1/8 pixel points are located between the first type of 1/8 pixel points and the third type of 1/8 pixel points 1/8 pixel, such as b2, b8, g2, g8. The filter of the seven-type 1/8 pixel interpolation unit 1212 has a tap coefficient of

{{0, 0, 3, 1, 0, 0}, {0, -1, -17, -2, 0, 0}, {3, -17, 237, 29, -6, 1}, {1 , -2, 29, 5, -2, 0}, {0, 0, -6, -1, 0, 0}, {0, 0, 1, 0, 0, 0}} 19-tap filter . in,

b2＝(3*A3+1*A4+(-1)*B2+(-17)*B3+(-2)*B4+3*C1+(-17)*C2+237*C3+29*C4+(-6) *C5+1*C6+1*D1+(-2)*D2+29*D3+5*D4+(-2)*D5+(-6)*E3+(-1)*E4+1*F3)/256.

b8＝(1*A3+3*A4+(-2)*B3+(-17)*B4+(-1)*B5+1*C1+(-6)*C2+29*C3+237*C4+(-17) *C5+3*C6+(-1)*D2+5*D3+29*D4+(-2)*D5+1*D6+(-2)*E3+(-6)*E4+1*F4)/256.

h2＝(1*A3+(-6)*B3+(-1)*B4+1*C1+(-2)*C2+29*C3+5*C4+(-2)*C5+3*D1+(-17) *D2+237*D3+29*D4+(-6)*D5+1*D6+(-1)*E2+(-17)*E3+(-2)*E4+3*F3+1*F4)/256.

h8＝(1*A4+(-2)*B3+(-6)*B4+(-1)*C2+5*C3+29*C4+(-2)*C5+1*C6+1*D1+(-6) *D2+29*D3+237*D4+(-17)*D5+3*D6+(-2)*E3+(-17)*E4+(-1)*E5+1*F3+3*F4)/256.

(13) Eight types of 1/8 pixel interpolation unit 1213 can interpolate to obtain eight types of 1/8 pixels. 1/8 pixels, 1/8 pixels between five types of 1/8 pixels, such as b4, b6, d2, d8, f2, f8, h4, h6. The filter of the eight types of 1/8 pixel interpolation unit 1213 is a tap coefficient of

{{0, 0, 2, 1, 0, 0}, {1, -1, -12, -7, -1, 0}, {5, -28, 182, 105, -20, 4}, { 1, -5, 24, 15, -5, 1}, {0, 0, -5, -3, 0, 0}, {0, 0, 1, 1, 0, 0}} 23-tap filtering device. in,

b4＝(2*A3+1*A4+1*B1+(-1)*B2+(-12)*B3+(-7)*B4+(-1)*B5+5*C1+(-28)*C2+182 *C3+105*C4+(-20)*C5+4*C6+1*D1+(-5)*D2+24*D3+15*D4+(-5)*D5+1*D6+(-5)*E3+ (-3)*E4+1*F3+1*F4)/256.

b6＝(1*A3+2*A4+(-1)*B2+(-7)*B3+(-12)*B4+(-1)*B5+1*B6+4*C1+(-20)*C2+105 *C3+182*C4+(-28)*C5+5*C6+1*D1+(-5)*D2+15*D3+24*D4+(-5)*D5+1*D6+(-3)*E3+ (-5)*E4+1*F3+1*F4)/256.

d2＝(1*A2+5*A3+1*A4+(-1)*B2+(-28)*B3+(-5)*B4+2*C1+(-12)*C2+182*C3+24*C4+ (-5)*C5+1*C6+1*D1+(-7)*D2+105*D3+15*D4+(-3)*D5+1*D6+(-1)*E2+(-20)*E3+ (-5)*E4+4*F3+1*F4)/256.

d8＝(1*A3+5*A4+1*A5+(-5)*B3+(-28)*B4+(-1)*B5+1*C1+(-5)*C2+24*C3+182*C4+ (-12)*C5+2*C6+1*D1+(-3)*D2+15*D3+105*D4+(-7)*D5+1*D6+(-5)*E3+(-20)*E4+ (-1)*E5+1*F3+4*F4)/256.

f2＝(4*A3+1*A4+(-1)*B2+(-20)*B3+(-5)*B4+1*C1+(-7)*C2+105*C3+15*C4+(-3) *C5+1*C6+2*D1+(-12)*D2+182*D3+24*D4+(-5)*D5+1*D6+(-1)*E2+(-28)*E3+(-5) *E4+1*F2+5*F3+1*F4)/256.

f8＝(1*A3+4*A4+(-5)*B3+(-20)*B4+(-1)*B5+1*C1+(-3)*C2+15*C3+105*C4+(-7) *C5+1*C6+1*D1+(-5)*D2+24*D3+182*D4+(-12)*D5+2*D6+(-5)*E3+(-28)*E4+(-1) *E5+1*F3+5*F4+1*F5)/256.

h4＝(1*A3+1*A4+(-5)*B3+(-3)*B4+1*C1+(-5)*C2+24*C3+15*C4+(-5)*C5+1*C6 +5*D1+(-28)*D2+182*D3+105*D4+(-20)*D5+4*D6+1*E1+(-1)*E2+(-12)*E3+(-7)*E4+ (-1)*E5+2*F3+1*F4)/256.

h6＝(1*A3+1*A4+(-3)*B3+(-5)*B4+1*C1+(-5)*C2+15*C3+24*C4+(-5)*C5+1*C6 +4*D1+(-20)*D2+105*D3+182*D4+(-28)*D5+5*D6+(-1)*E2+(-7)*E3+(-12)*E4+(-1) *E5+1*E6+1*F3+2*F4)/256.

(14) Nine types of 1/8 pixel interpolation unit 1214 can interpolate to obtain nine types of 1/8 pixels, and nine types of 1/8 pixels are located between five types of 1/8 pixels and six types of 1/8 pixels 1/8 pixel, such as d4, d6, f4, f6. The filter of the nine types of 1/8 pixel interpolation unit 1214 is a tap coefficient of

{{0, 0, 4, 3, 0, 0}, {0, 0, -21, -13, 0, 0}, {4, -21, 141, 82, -16, 3}, {2, -13, 82, 48, -10, 2}, {0, 0, -16, -10, 0, 0}, {0, 0, 3, 2, 0, 0}} 20-tap filters. in,

d4＝(4*A3+3*A4+(-21)*B3+(-13)*B4+4*C1+(-21)*C2+141*C3+82*C4+(-16)*C5+3*C6 +2*D1+(-13)*D2+82*D3+48*D4+(-10)*D5+2*D6+(-16)*E3+(-10)*E4+3*F3+2*F4)/ 256.

d6＝(3*A3+4*A4+(-13)*B3+(-21)*B4+3*C1+(-16)*C2+82*C3+141*C4+(-21)*C5+4*C6 +2*D1+(-10)*D2+48*D3+82*D4+(-13)*D5+2*D6+(-10)*E3+(-16)*E4+2*F3+3*F4)/ 256.

f4＝(3*A3+2*A4+(-16)*B3+(-10)*B4+2*C1+(-13)*C2+82*C3+48*C4+(-10)*C5+2*C6 +4*D1+(-21)*D2+141*D3+82*D4+(-16)*D5+3*D6+(-21)*E3+(-13)*E4+4*F3+3*F4)/ 256.

f6＝(2*A3+3*A4+(-10)*B3+(-16)*B4+2*C1+(-10)*C2+48*C3+82*C4+(-13)*C5+3*C6 +3*D1+(-16)*D2+82*D3+141*D4+(-21)*D5+4*D6+(-13)*E3+(-21)*E4+2*F3+4*F4)/ 256.

The output module 130 is respectively connected to the above-mentioned interpolation units, and outputs the pixel values of the sub-pixels obtained by the above-mentioned interpolation.

According to the 1/8 pixel precision interpolation device of the embodiment of the present invention, for all sub-pixel position interpolation, only integer pixels need to be used as reference pixels, which simplifies the interpolation process in sub-pixel positions, thereby reducing the time delay of the interpolation process. At the same time, the interpolation device of the embodiment of the present invention adopts a high-precision interpolation filter, which can make the motion vector accurate to 1/8 pixel level in motion compensation, improve the prediction accuracy, reduce the code rate, and improve the compression efficiency of video coding.

In the description of this specification, descriptions referring to the terms "one embodiment", "some embodiments", "example", "specific examples", or "some examples" mean that specific features described in connection with the embodiment or example , structure, material or characteristic is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

Although the embodiments of the present invention have been shown and described, those skilled in the art can understand that various changes, modifications and substitutions can be made to these embodiments without departing from the principle and spirit of the present invention. and modifications, the scope of the invention is defined by the appended claims and their equivalents.

Claims (11)

1. A 1/8 pixel precision interpolation method, is characterized in that, comprises the steps: Set the sub-pixel category and the position of the sub-pixel, the sub-pixel includes: half pixel, 1/4 pixel and 1/8 pixel, Wherein, the half-pixels include one-type half-pixels and second-type half-pixels, the first-type half-pixels are half-pixels in the same row or column as the integer pixels, and the second-type half-pixels The pixels are the remaining half-pixels in the half-pixels; The 1/4 pixels include one type of 1/4 pixels, two types of 1/4 pixels, and three types of 1/4 pixels, and the one type of 1/4 pixels is the same as the integer pixel A 1/4 pixel in a row or the same column, the second type of 1/4 pixel point is in the same row with the first type of half pixel point and the second type of half pixel point at the same time or simultaneously with the first type of half pixel point and the second type of half pixel point The 1/4 pixel points of the same half-pixel point are in the same column, and the three types of 1/4 pixel points are the remaining 1/4 pixel points in the 1/4 pixel points; The 1/8 pixel points include class 1/8 pixel points, class 2 1/8 pixel points, class 3 1/8 pixel points, class 4 1/8 pixel points, class 5 1/8 pixel points, class 6 1/8 pixels, seven types of 1/8 pixels, eight types of 1/8 pixels, and nine types of 1/8 pixels, the type 1/8 pixels are located between the integer pixels and the type 1/8 1/8 pixels between 4 pixels; the second type of 1/8 pixels are 1/8 pixels between the first type of half pixels and the first type of 1/4 pixels, and the three Class 1/8 pixel points are 1/8 pixel points located between the first class 1/4 pixel points and third class 1/4 pixel points, and the four class 1/8 pixel points are 1/8 pixel points located in the first class and a half The 1/8 pixel point between the pixel point and the second type 1/4 pixel point, the fifth type 1/8 pixel point is the 1/8 pixel point between the third type 1/4 pixel point and the second type 1/4 pixel point 8 pixels, the six types of 1/8 pixels are 1/8 pixels between the second type of 1/4 pixels and the second type of half pixels, and the seven types of 1/8 pixels are located The 1/8 pixels between the 1/8 pixels of the first category and the 1/8 pixels of the third category, the 1/8 pixels of the eight categories are located in the 1/8 pixels of the second category, the 1/8 pixels of the third category /8 pixels, four types of 1/8 pixels, 1/8 pixels between five types of 1/8 pixels, the nine types of 1/8 pixels are located in the five types of 1/8 pixels and six types of pixels 1/8 pixels between 1/8 pixels; and Select the integer pixel in the (M+5)×(N+5) integer pixel block closest to the sub-pixel block formed by the sub-pixel from the original reference frame as the reference pixel, according to the set various types The position of the sub-pixels is interpolated to obtain the sub-pixels, wherein M is the number of sub-pixels in each row in the sub-pixel block to be obtained by interpolation, and N is the number of sub-pixels in each row of the sub-pixels to be obtained by interpolation. The number of pixels in the column. 2. The interpolation method according to claim 1, characterized in that, according to the positions of various sub-pixels that have been set, use interpolation filters to obtain various sub-pixels, wherein the interpolation filter corresponding to each class of sub-pixels The sum of the tap coefficients is 256. 3. The interpolation method according to claim 1, wherein the number of sub-pixel points in each row in the sub-pixel block is 1, and the number of sub-pixel points in each column is 1. 4. The interpolation method according to claim 3, wherein the number of taps of the interpolation filter is less than 36. 5. The interpolation method according to claim 4, characterized in that, the tap coefficients of said one type of half pixel points are {{0,0,0,0,0,0}, {1,-3,2, 2, -3, 1}, {5, -27, 150, 150, -27, 5}, {1, -3, 2, 2, -3, 1}, {0, 0, 0, 0, 0 , 0}, 18-tap filter interpolation acquisition of {0, 0, 0, 0, 0, 0, 0}}; The second type of half-pixel points through the tap coefficient is {{0, 0, 3, 3, 0, 0}, {0, 0, -16, -16, 0, 0}, {3, -16, 90, 90, -16, 3}, {3, 20-tap filter interpolation of -16, 90, 90, -16, 3}, {0, 0, -16, -16, 0, 0}, {0, 0, 3, 3, 0, 0}} Obtain; The one-type 1/4 pixel points are passed through the tap coefficient as {{0, 0, 0, 0, 0, 0}, {1, -2, 1, 1, -2, 1}, {5, -27, 223, 67, -14, 2}, {1, -2, 1, 1, -2, 1}, {0, 0, 0, 0, 0, 0}, {0, 0, 0, 0, 0, 0}} 18-tap filter interpolation acquisition; The second type of 1/4 pixels pass the tap coefficients as {{0, 0, 3, 3, 0, 0}, {0, 0, -15, -15, 0, 0}, {4, -23, 132, 132, -23, 4}, {2, -9, 41, 41, -9, 2} {0, 0, -9, -9, 0, 0}, {0, 0, 2, 2, 0, 0}} 20-tap filter interpolation acquisition; The three types of 1/4 pixel points pass the tap coefficient as {{0, 0, 4, 2, 0, 0}, {0, 0, -23, -8, 0, 0}, {4, -23, 194, 60, -12, 2}, {2, 20-tap filter interpolation of -8, 60, 20, -5, 1}, {0, 0, -12, -5, 0, 0}, {0, 0, 2, 1, 0, 0}} Obtain; Said one type of 1/8 pixels is passed through the tap coefficient as {{0, 0, 0, 0, 0, 0}, {0, -1, 1, 1, -1, 0}, {3, -18, 246, 30, -6, 1}, {0, -1, 1, 1, -1, 0}, {0, 0, 0, 0, 0, 0}, {0, 0, 0, 0, 0, 0}} 14-tap filter interpolation acquisition; The second type of 1/8 pixel points through the tap coefficient is {{0, 0, 0, 0, 0, 0}, {1, -3, 2, 1, -2, 1}, {5, -29, 189, 108, -21, 4}, {1, -3, 2, 1, -2, 1}, {0, 0, 0, 0, 0, 0}, {0, 0, 0, 0, 0, 0}} 18-tap filter interpolation acquisition; The three types of 1/8 pixels pass through the tap coefficient as {{0, 0, 3, 1, 0, 0}, {1, -1, -14, -4, -1, 0}, {4, -25, 214, 65, -13, 2}, { 1, -5, 27, 9, -3, 1}, {0, 0, -6, -2, 0, 0} {0, 0, 1, 1, 0, 0}} 22-tap filter interpolation acquisition; The four types of 1/8 pixel points pass the tap coefficient as {{0, 0, 2, 2, 0, 0}, {1, -1, -9, -9, -1, 1}, {4, -26, 145, 145, -26, 4}, { 1, -5, 19, 19, -5, 1}, {0, 0, -4, -4, 0, 0}, {0, 0, 1, 1, 0, 0}} 24-tap filtering The device interpolation acquisition; The tap coefficients of the five types of 1/8 pixels are {{0, 0, 4, 2, 0, 0}, {0, 0, -19, -11, 0, 0}, {4, -25, 166, 95, -18, 3}, {2, 20-tap filter interpolation of -9, 51, 31, -7, 2}, {0, 0, -11, -7, 0, 0}, {0, 0, 2, 1, 0, 0}} Obtain; The six types of 1/8 pixel points pass the tap coefficient as {{0, 0, 3, 3, 0, 0}, {0, 0, -17, -17, 0, 0}, {4, -20, 113, 113, -20, 4}, {2, 20-tap filter interpolation of -12, 66, 66, -12, 2}, {0, 0, -13, -13, 0, 0}, {0, 0, 2, 2, 0, 0}} Obtain; The seven types of 1/8 pixel points pass the tap coefficient as {{0, 0, 3, 1, 0, 0}, {0, -1, -17, -2, 0, 0}, {3, -17, 237, 29, -6, 1}, {1 , -2, 29, 5, -2, 0}, {0, 0, -6, -1, 0, 0}, {0, 0, 1, 0, 0, 0}} 19-tap filter interpolation acquisition; The eight types of 1/8 pixel points pass the tap coefficient as {{0, 0, 2, 1, 0, 0}, {1, -1, -12, -7, -1, 0}, {5, -28, 182, 105, -20, 4}, { 1, -5, 24, 15, -5, 1}, {0, 0, -5, -3, 0, 0}, {0, 0, 1, 1, 0, 0}} 23-tap filtering The device interpolation acquisition; The nine types of 1/8 pixel points pass the tap coefficient as {{0, 0, 4, 3, 0, 0}, {0, 0, -21, -13, 0, 0}, {4, -21, 141, 82, -16, 3}, {2, 20-tap filter interpolation of -13, 82, 48, -10, 2}, {0, 0, -16, -10, 0, 0}, {0, 0, 3, 2, 0, 0}} Obtain. 6. A 1/8 pixel precision interpolation device, characterized in that, comprising: An acquisition module, configured to acquire an integer pixel as a reference pixel from the original reference frame; A sub-pixel interpolation module, the sub-pixel interpolation module is connected to the acquisition module, and is used to receive integer pixels from the acquisition module and perform interpolation on the integer pixels to obtain sub-pixels, the sub-pixels Including: half pixel, 1/4 pixel and 1/8 pixel, the sub-pixel interpolation module includes: A type of half-pixel interpolation unit, used for interpolation to obtain a type of half-pixel, where the type of half-pixel is a half-pixel in the same row or column as the integer pixel; a second-type half-pixel interpolation unit, configured to obtain second-type half-pixels by interpolation, and the second-type half-pixels are remaining half-pixels in the half-pixels; A type of 1/4 pixel interpolation unit, used for interpolation to obtain a type of 1/4 pixel, the type of 1/4 pixel is a 1/4 pixel in the same row or column as the integer pixel; The second-type 1/4 pixel interpolation unit is used for interpolation to obtain the second-type 1/4 pixel point, and the second-type 1/4 pixel point is in the same row as the first-type half-pixel point and the second-type half-pixel point at the same time or At the same time, the 1/4 pixel points in the same column as the first-type half-pixel point and the second-type half-pixel point; Three types of 1/4 pixel interpolation unit, used for interpolation to obtain three types of 1/4 pixels, the three types of 1/4 pixels are the remaining 1/4 pixels in the 1/4 pixels; A type of 1/8 pixel interpolation unit, used for interpolation to obtain a type of 1/8 pixel point, the type of 1/8 pixel point is 1/8 between the integer pixel and a type of 1/4 pixel point pixel; The second type of 1/8 pixel interpolation unit is used for interpolation to obtain the second type of 1/8 pixel points, the second type of 1/8 pixel points are located between the first type of half pixel points and the first type of 1/4 pixel points 1/8 pixel; Three types of 1/8 pixel interpolation unit, used for interpolation to obtain three types of 1/8 pixels, the three types of 1/8 pixels are located between the one type of 1/4 pixels and the three types of 1/4 pixels 1/8 pixels between; Four types of 1/8 pixel interpolation units, used for interpolation to obtain four types of 1/8 pixel points, the four types of 1/8 pixel points are located between the first type of half pixel points and the second type of 1/4 pixel points 1/8 pixel; Five types of 1/8 pixel interpolation unit, used for interpolation to obtain five types of 1/8 pixels, the five types of 1/8 pixels are located between the third type of 1/4 pixels and the second type of 1/4 pixels 1/8 pixel; Six types of 1/8 pixel interpolation units, used for interpolation to obtain six types of 1/8 pixel points, the six types of 1/8 pixel points are located between the second type of 1/4 pixel points and the second type of half pixel points 1/8 pixel; Seven types of 1/8 pixel interpolation unit, used for interpolation to obtain seven types of 1/8 pixel points, the seven types of 1/8 pixel points are located between the first type of 1/8 pixel points and the three types of 1/8 pixel points 1/8 pixels between; Eight types of 1/8 pixel interpolation unit, used for interpolation to obtain eight types of 1/8 pixel points, the eight types of 1/8 pixel points are located in the second type of 1/8 pixel points, the third type of 1/8 pixel points, Four types of 1/8 pixels, 1/8 pixels between five types of 1/8 pixels; Nine types of 1/8 pixel interpolation unit, used for interpolation to obtain nine types of 1/8 pixel points, the nine types of 1/8 pixel points are located between the fifth type of 1/8 pixel points and the sixth type of 1/8 pixel points 1/8 pixel; and An output module, the output module is connected to the sub-pixel interpolation module, and is used to output sub-pixel values of sub-pixel points obtained by interpolation. 7. The interpolation device according to claim 6, wherein the integer pixel as the reference pixel is the closest (M+5)×( N+5) Integer pixel points in the integer pixel block, wherein M is the number of sub-pixel points in each row in the sub-pixel block to be obtained by interpolation, and N is each column in the described sub-pixel points to be obtained by interpolation The number of sub-pixels. 8. The interpolation device according to claim 7, wherein the number of sub-pixel points in each row in the sub-pixel block is 1, and the number of sub-pixel points in each column is 1. 9. The interpolation device according to claim 8, wherein the first type of half pixel interpolation unit, the second type of half pixel interpolation unit, the first type of 1/4 pixel interpolation unit, the second type of 1/4 pixel interpolation unit, Three types of 1/4 pixel interpolation unit, one type of 1/8 pixel interpolation unit, two types of 1/8 pixel interpolation unit, three types of 1/8 pixel interpolation unit, four types of 1/8 pixel interpolation unit, five types of 1/8 The pixel interpolation unit, six types of 1/8 pixel interpolation units, seven types of 1/8 pixel interpolation units, eight types of 1/8 pixel interpolation units and nine types of 1/8 pixel interpolation units are filters. 10. The interpolation device according to claim 9, wherein the first type of half pixel interpolation unit, the second type of half pixel interpolation unit, the first type of 1/4 pixel interpolation unit, the second type of 1/4 pixel interpolation unit, Three types of 1/4 pixel interpolation unit, one type of 1/8 pixel interpolation unit, two types of 1/8 pixel interpolation unit, three types of 1/8 pixel interpolation unit, four types of 1/8 pixel interpolation unit, five types of 1/8 The sum of the tap coefficients of the filters of the pixel interpolation unit, six types of 1/8 pixel interpolation units, seven types of 1/8 pixel interpolation units, eight types of 1/8 pixel interpolation units and nine types of 1/8 pixel interpolation units is 256 , and the number of taps of each filter is less than 36. 11. The interpolation device according to claim 10, characterized in that, the filter of said type of half-pixel interpolation unit has tap coefficients {{0, 0, 0, 0, 0, 0}, {1,- 3, 2, 2, -3, 1}, {5, -27, 150, 150, -27, 5}, {1, -3, 2, 2, -3, 1}, {0, 0, 0 , 0, 0, 0}, {0, 0, 0, 0, 0, 0, 0}} is an 18-tap filter; The filter of the second type of half-pixel interpolation unit has a tap coefficient of {{0, 0, 3, 3, 0, 0}, {0, 0, -16, -16, 0, 0}, {3, -16, 90, 90, -16, 3}, {3, -16, 90, 90, -16, 3}, {0, 0, -16, -16, 0, 0}, {0, 0, 3, 3, 0, 0}} 20-tap filter; The filter of the second type of half-pixel interpolation unit has a tap coefficient of {{0, 0, 3, 3, 0, 0}, {0, 0, -16, -16, 0, 0}, {3, -16, 90, 90, -16, 3}, {3, -16, 90, 90, -16, 3}, {0, 0, -16, -16, 0, 0}, {0, 0, 3, 3, 0, 0}} 20-tap filter The filter of the second type 1/4 pixel interpolation unit has tap coefficients {{0, 0, 3, 3, 0, 0}, {0, 0, -15, -15, 0, 0}, {4 , -23, 132, 132, -23, 4}, {2, -9, 41, 41, -9, 2} {0, 0, -9, -9, 0, 0}, {0, 0, 2, 2, 0, 0}} 20-tap filter; The filters of the three types of 1/4 pixel interpolation units have a tap coefficient of {{0, 0, 4, 2, 0, 0}, {0, 0, -23, -8, 0, 0}, {4, -23, 194, 60, -12, 2}, {2, 20-tap filters of -8, 60, 20, -5, 1}, {0, 0, -12, -5, 0, 0}, {0, 0, 2, 1, 0, 0}}; The filter of the described class 1/8 pixel interpolation unit is that the tap coefficient is {{0, 0, 0, 0, 0, 0}, {0, -1, 1, 1, -1, 0}, {3, -18, 246, 30, -6, 1}, {0, -1, 1, 1, -1, 0}, {0, 0, 0, 0, 0, 0}, {0, 0, 0, 0, 0, 0}} 14-tap filter; The filter of the second type 1/8 pixel interpolation unit is that the tap coefficient is {{0, 0, 0, 0, 0, 0}, {1, -3, 2, 1, -2, 1}, {5, -29, 189, 108, -21, 4}, {1, -3, 2, 1, -2, 1}, {0, 0, 0, 0, 0, 0}, {0, 0, 0, 0, 0, 0}} 18-tap filter; The filters of the three types of 1/8 pixel interpolation units are tap coefficients of {{0, 0, 3, 1, 0, 0}, {1, -1, -14, -4, -1, 0}, {4, -25, 214, 65, -13, 2}, { 1, -5, 27, 9, -3, 1}, {0, 0, -6, -2, 0, 0} {0, 0, 1, 1, 0, 0}} 22-tap filter ; The filters of the four types of 1/8 pixel interpolation units are tap coefficients of {{0, 0, 2, 2, 0, 0}, {1, -1, -9, -9, -1, 1}, {4, -26, 145, 145, -26, 4}, { 1, -5, 19, 19, -5, 1}, {0, 0, -4, -4, 0, 0}, {0, 0, 1, 1, 0, 0}} 24-tap filtering device; The filter of the five types of 1/8 pixel interpolation units is that the tap coefficient is {{0, 0, 4, 2, 0, 0}, {0, 0, -19, -11, 0, 0}, {4, -25, 166, 95, -18, 3}, {2, 20-tap filters of -9, 51, 31, -7, 2}, {0, 0, -11, -7, 0, 0}, {0, 0, 2, 1, 0, 0}}; The filter of the six types of 1/8 pixel interpolation units is that the tap coefficient is {{0, 0, 3, 3, 0, 0}, {0, 0, -17, -17, 0, 0}, {4, -20, 113, 113, -20, 4}, {2, -12, 66, 66, -12, 2}, {0, 0, -13, -13, 0, 0}, {0, 0, 2, 2, 0, 0}} 20-tap filter; The filter of the seven types of 1/8 pixel interpolation units is that the tap coefficient is {{0, 0, 3, 1, 0, 0}, {0, -1, -17, -2, 0, 0}, {3, -17, 237, 29, -6, 1}, {1 , -2, 29, 5, -2, 0}, {0, 0, -6, -1, 0, 0}, {0, 0, 1, 0, 0, 0}} 19-tap filter ; The filter of the eight types of 1/8 pixel interpolation units is that the tap coefficient is {{0, 0, 2, 1, 0, 0}, {1, -1, -12, -7, -1, 0}, {5, -28, 182, 105, -20, 4}, { 1, -5, 24, 15, -5, 1}, {0, 0, -5, -3, 0, 0}, {0, 0, 1, 1, 0, 0}} 23-tap filtering device; The filter of the nine types of 1/8 pixel interpolation units is that the tap coefficient is {{0, 0, 4, 3, 0, 0}, {0, 0, -21, -13, 0, 0}, {4, -21, 141, 82, -16, 3}, {2, -13, 82, 48, -10, 2}, {0, 0, -16, -10, 0, 0}, {0, 0, 3, 2, 0, 0}} 20-tap filters.

Images