JP3046689B2 - Motion vector detection method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method of detecting a motion vector, and more particularly to a method of detecting a motion vector by block matching between a prediction target image and a reference image when coding a moving image using motion compensated prediction. The present invention relates to a vector detection method .

[0002]

2. Description of the Related Art As a main method for obtaining a motion vector by block matching, a document "Interface Au" is used.
g. 1992 pp138 ", a full search, a three-step search, and the like. Hereinafter, the calculation amount of the conventional motion vector detection method will be described.

Here, sub is subtraction, add is addition, o
ps indicates the amount of calculation. Also, the size of the block is 16 ×
It is assumed that 16 rectangular blocks, the search range of the motion vector is a rectangular area of −15 to +15 in both the horizontal and vertical directions, and the accuracy of the motion vector is in units of one pixel. The sum of the values shall be used.

[0004] 1. Full Search (1) A process of obtaining a prediction error PE using all pixels in a block is performed for all candidate blocks of 961 (31 ² ).

(2) A motion vector for a candidate block having a minimum prediction error PE is determined as an optimum motion vector.

[0006] In the above processing, the amount of calculation required for block matching is as follows: Calculation of prediction error PE: 256 (sub) +255 (ad)
d) = 511 ops Comparison between prediction error PE and minimum value: Since 1 ops is performed for all 961 candidate blocks, (511 + 1) × 961 = 492032 ops.

2.3 Step Search (1) A process of obtaining a prediction error PE using all pixels in a block is performed on a candidate block corresponding to a motion vector (4m, 4n). Here, m and n are both -3.
It is an integer of up to +3.

(2) The minimum value of the prediction error PE in (1) is MIN_PE, the motion vector giving MIN_PE is (x, y), and the motion vector is (x-2, y-2);
(X, y-2), (x + 2, y-2), (x-2,
y), (x + 2, y), (x−2, y + 2), (x, y)
+2) and eight candidate blocks corresponding to (x + 2, y + 2), a prediction error P using all pixels in the block
E is obtained, and if PE <MIN_PE, MIN_PE =
PE.

(3) The motion vector giving MIN_PE in (2) is (x, y), and the motion vector is (x-1, y).
-1), (x, y-1), (x + 1, y-1), (x-
1, x), (x + 1, y) (x-1, y + 1), (x,
For each of eight candidate blocks corresponding to (y + 1) and (x + 1, y + 1), a prediction error PE using all pixels in the block is obtained. If PE <MIN_PE, MIN_PE
= PE.

(4) The motion vector giving MIN_PE in (3) is set as the optimum motion vector.

The amount of calculation required for block matching in the above processing is as follows: Calculation of prediction error PE: 511 ops Comparison between prediction error PE and minimum value MIN_PE: 1 ops 1) Number of times to obtain prediction error PE by 7) = 7 × 7 = 49 times The number of times to calculate the prediction error PE in 2): 8 times 3) The number of times to calculate the prediction error PE: 8 times (511 + 1) × (49 + 8 + 8) = 3328
It becomes 0 ops.

3.2 Step Search (1) A process of obtaining a prediction error PE using all pixels in a block is performed on a candidate block corresponding to a motion vector (3m, 3n). Here, m and n are both-
It is an integer of 5 to +5.

(2) The minimum value of the prediction error PE in (1) is MIN_PE, the motion vector to which MIN_PE is given is (x, y), and the motion vector (x-1, y-
1), (x, y-1), (x + 1, y-1), (x-
1, y), (x + 1, y), (x-1, y + 1),
For eight candidate blocks corresponding to (x, y + 1) and (x + 1, y + 1), a prediction error PE using all pixels in the block is obtained. If PE <MIN_PE, MIN is obtained.
_PE = PE.

(3) The motion vector giving MIN_PE in (2) is set as the optimum motion vector.

The amount of calculation required for block matching in the above processing is as follows: Calculation of prediction error PE: 511 ops Comparison between prediction error PE and minimum value MIN_PE: 1 ops 1) Number of times of obtaining prediction error PE by 11) = 11 × 11 = 121
The number of times the prediction error PE is calculated in the second time 2): 8 times, so that (511 + 1) × (121 + 8) = 66048
ops.

[0016]

However, full search requires an enormous amount of calculation for block matching as described above. Also, in the three-step search and the two-step search, since only a part of the candidate blocks are checked, a correct motion block cannot be obtained, and the image quality of the reproduced image is significantly inferior to that in the case of the full search.

An object of the present invention is to provide a motion vector detecting method capable of obtaining a correct motion vector even if the amount of calculation is greatly reduced.

[0018]

SUMMARY OF THE INVENTION The present invention provides a motion picture detecting method for detecting a motion vector by performing block matching between a picture to be predicted consisting of L pixels and a reference picture in moving picture coding using motion compensated prediction. A vector detecting method, comprising: a first step of calculating a first prediction error for L1 (L1 <L) pixels by block matching between a candidate block and a prediction target image; A second step of calculating a second prediction error by block matching between a candidate block and a prediction target image for L2 (L1 <L2 ≦ L) pixels only when the threshold T is equal to or less than the threshold T. if the second prediction error second prediction error is smallest, and a third step of the motion vector and the motion vector of the prediction target image, each candidate block
The first to third steps are performed for every candidate.
It is characterized in that a motion vector is obtained by repeating until block matching for a block is completed.

[0019]

According to the present invention, in a moving picture encoding using motion compensated prediction, a prediction target picture composed of L pixels
A motion vector detection method for detecting a motion vector by performing block matching with a reference image, wherein L1 (L1
A first step of calculating a first prediction error for the <L) pixels by block matching between the candidate block and the image to be predicted, and L2 only when the first prediction error is equal to or less than a predetermined threshold T. A second step of calculating a second prediction error for the (L1 <L2 ≦ L) pixels by block matching between the candidate block and the prediction target image, and a second step in which the obtained second prediction error is the minimum. If it is a prediction error, the third step of setting the motion vector as the motion vector of the prediction target image is performed for each candidate block ,
Perform the first to third steps for all candidate blocks.
Since the motion vector is repeatedly obtained until block matching is completed, the amount of calculation and memory required for block matching can be significantly reduced, and a highly accurate motion vector can be detected.

[0020]

An embodiment will be described below with reference to the drawings. In each case, a block composed of L pixels is used as a unit, and block matching is performed between one block in the prediction target image and N candidate blocks in the reference image to detect a motion vector. I have.

FIG. 1 is a diagram showing an example of a circuit for realizing a first embodiment of the motion vector detecting method according to the present invention.

In FIG. 1, reference numeral 1 denotes an address generator, and L in a unit block in response to a control signal from a control unit 9.
Address information for a pixel used to determine a prediction error PE1 for 1 (<L) pixels, and L2
Prediction error PE2 for (L1 <L2 ≦ L) pixels
Generates address information for the pixels used in the calculation.

Here, the L2 pixels used in obtaining the prediction error PE2 are L2 pixels used in obtaining the prediction error PE1.
When calculating the prediction error PE2, including only one pixel, only the address information for the (L2-L1) pixels that are not used when calculating the prediction error PE1 out of the L2 pixels is generated.

An address generator 2 receives a control signal from the control unit 9 and receives a prediction error P in a candidate block corresponding to the motion vector information from the motion vector generation unit 10.
The address information for the L1 pixels used for obtaining E1 and the address information for the L2 pixels used for obtaining the prediction error PE2 are generated.

Here, the L2 pixels used for obtaining the prediction error PE2 include the L1 pixels used for obtaining the prediction error PE1, and the L2 pixels in the candidate block are used for obtaining the prediction error PE2. Among them, only the address information for the (L2-L1) pixels not used in obtaining the prediction error PE1 is generated.

Reference numeral 3 denotes a memory for storing the information of the image to be predicted in units of blocks, from which the stored information is read in accordance with the information from the address generator 1.

Reference numeral 4 denotes a memory for storing information on a search area in the reference image, from which information is read in accordance with information from the address generator 2.

Reference numeral 5 denotes a prediction error calculation unit for obtaining a prediction error between a prediction target block and a candidate block. When the prediction error PE2 is obtained, the prediction errors for the (L2−L1) pixels that are not used in obtaining the prediction error PE1 among the L2 pixels are obtained, and the value is calculated as the prediction error P
It is determined by adding to E1. As a method of calculating the prediction error, for example, a cumulative sum of absolute values of pixel value differences between corresponding pixels of the prediction target block and the candidate block can be used.

Reference numeral 6 denotes a comparing section for comparing the prediction error PE1 with the threshold value TH.

Reference numeral 7 denotes a comparison unit. Reference numeral 8 denotes a register for storing the minimum value MIN_PE of the prediction error PE2 and a motion vector for giving the minimum value MIN_PE. The comparison unit 7 stores the prediction error PE2 and the minimum value stored in the register 8. Compare MIN_PE.

Reference numeral 9 denotes a control unit, and as a result of the processing of the comparison unit 6, when the prediction error PE1 is equal to or less than the threshold value TH,
A control signal for obtaining the prediction error PE2 is generated, and as a result of the processing of the comparing section 6, when the prediction error PE1 is larger than the threshold value TH and when the processing of the comparing section 7 is completed, the next candidate block is generated. To generate a control signal for obtaining the prediction error PE1.

Reference numeral 10 denotes a motion vector generation unit which is controlled by the control unit 9 when the prediction error PE1 is larger than the threshold value TH as a result of the processing of the comparison unit 6 and when the processing of the comparison unit 7 is completed. Generate motion vector information for the next candidate block.

In the circuit described above, the address generators 1,
2, the memories 3 and 4, the prediction error calculation unit 5, the comparison unit 6, the control unit 9 and the motion vector generation unit 10 constitute a calculation unit, and the comparison unit 7, the register 8, the control unit 9 and the motion vector generation unit 10 The comparison means is constituted.

The processing procedure will be described below.

(1) The threshold value of the prediction error PE1 is set to TH.

(2) The prediction error calculator 5 calculates L
The prediction error PE1 is obtained using 1 (<L) pixels, and the comparison unit 6 compares the prediction error PE1 with the threshold value TH.

(3) If the prediction error PE1 is equal to or smaller than the threshold value, the prediction error calculation unit 5 sets L2 (L1 <L
The prediction error PE2 is obtained using 2 ≦ L) pixels, and the comparison unit 7 compares the prediction error PE2 with the minimum value MIN_PE. If the prediction error PE2 is smaller than the minimum value MIN_PE, the prediction error PE2 is newly minimized. It is stored in the register 8 together with the motion vector as the value MIN_PE.

(4) The processes (2) and (3) are performed on all candidate blocks, and the motion vector stored in the register 8 at the end is determined as the optimum motion vector.

As the initial setting of the above processing, the minimum value MIN_PE of the prediction error PE2 is set in the register 8 as the maximum possible value of the prediction error PE2. When the prediction error PE1 is larger than the threshold value TH in all the candidate blocks, the means for determining a motion vector include: (1) using a specific motion vector; and (2) the same motion vector as the immediately preceding prediction target block. (3) There is a method in which a motion vector to a candidate block with a minimum prediction error PE1 is set as an optimal motion vector.

The method (1) can be performed by setting a specific value such as (0, 0) in the register 8 as a motion vector in the initial setting.

The method (2) can be performed by holding the value of the immediately preceding process without setting the motion vector of the register 8 in the initial setting.

The method (3) can be performed by providing means for storing the minimum value of the prediction error PE1 and the value of the motion vector at that time.

FIG. 7 is a diagram for explaining means for obtaining the minimum value of the prediction error PE1. In the figure, 77 is a comparison unit,
A register 78 compares the prediction error PE1 with the minimum value of the prediction error PE1 stored in the register 8, and the register 78 stores the minimum value of the prediction error PE1 and a motion vector corresponding thereto. . Register 78
Is the prediction error PE in which the prediction error PE1 has already been stored.
It is updated when it is smaller than the minimum value of 1. As described above, the method (3) can be performed by adding the comparison unit 77 and the register 78 to the prediction error calculation unit 5.

The amount of calculation required for block matching in the above method will be described. Here, sub indicates subtraction, add indicates addition, and ops indicates the amount of calculation. In addition, a rectangular block having a block size of 16 × 16, and a search range of a motion vector is −15 to +1 in both the horizontal direction and the vertical direction.
It is assumed that the accuracy of the five rectangular areas and the motion vectors is in units of one pixel.

Assuming that L1 = 16 and L2 = 256, the amount of calculation required in the above processing is: calculation of prediction error PE1: 16 (sub) +15 (ad
d) = 31 ops Comparison of prediction error PE1 with threshold value TH: 1 ops Calculation of prediction error PE2: 240 (sub) +240 (a)
dd) = 480 ops Comparison between the prediction error PE2 and the minimum value MIN_PE: 1 ops. Therefore, assuming that the number of candidate blocks for which the prediction error PE2 is obtained is α, (31 + 1) × 31 ² + (480+
1) × α = 30752 + 481αops.

As described above, the prediction error PE1 is set only for the candidate block whose prediction error PE1 is equal to or smaller than the threshold value TH.
By obtaining 2, a motion vector can be obtained with a smaller amount of calculation than in a full search.

FIG. 6 is a diagram for explaining pixels used when calculating a prediction error for L1 pixels. The black squares in FIG. 6 indicate pixels used when calculating a prediction error for L1 (16) pixels.

The case where the above equation is applied to a motion vector detecting portion of the moving picture coding system MPEG2, which is currently being standardized by the ISO, will be described. In determining the prediction error PE1, 16 pixels shown in FIG. 6 were used, and the threshold value TH was set to 320. When the prediction errors PE1 of all the candidate blocks are larger than the threshold value TH, the motion vector to the candidate block with the smallest prediction error PE1 is determined as the optimal motion vector.

The sequence 1 is a sequence 2 with panning.
A second moving image, Sequence 2, is a two-second moving image in which an object moving in the horizontal and vertical directions is present.

(1) PE1> TH and the ratio of candidate blocks deleted is (unit:%).
0.09, and 80.28 in sequence 2.

(2) The average value (unit: pixel) of the difference from the motion vector obtained by the full search is as follows.

FD indicates the distance between the frame to be predicted and the reference frame. Not only this example but also 3
The results of the step search and the two-step search are also described.

Sequence 1 FD 3 step search 2 step search This embodiment XYXYXY 1 1.21 1.40 0.97 0.83 0.08 0.09 2 1.59 1.38 1.43 1.13 0.25 0.25 3 2.20 1.89 1.84 1.47 0.52 0.45 Sequence 2 FD 3 step search 2 step Search This embodiment XYXYXY1 1.50 1.38 1.60 1.36 0.11 0.10 2 1.83 1.36 1.18 0.89 0.18 0.15 3 2.17 1.80 1.91 1.43 0.27 0.20 (3) Image quality (SNR, unit: d) of the luminance signal of the reproduced image
b) A three-step search for reference as well as this embodiment,
The results of the step search are also shown. The value in parentheses is the difference from the full search.

Sequence 1 Full Search 3 Step Search 2 Step Search Example 28.20 27.14 27.30 27.90 (−0.88) (−0.72) (−0.12) Sequence 2 Full Search 3-step search 2-step search This embodiment 25.96 24.57 24.99 25.85 (-1.39) (-0.97) (-0.11) From the above results, in this embodiment, full search and Even if the calculation amount is reduced from 70% to 80%, the image quality (SN
It can be seen that almost the same motion vector can be obtained also at the point R). Further, as compared with the three-step search and the two-step search, the motion vector can be obtained more accurately and the image quality is good.

FIG. 2 is a diagram showing an example of a circuit for realizing a second embodiment of the motion vector detecting method according to the present invention. A counter 21 and a comparison unit 22 are added to the circuit of FIG. The counter 21 counts the number of candidate blocks for which the prediction error PE1 is equal to or smaller than the threshold value TH. The comparison unit 22 includes a counter 21
Is compared with the value of the maximum value M (<the total number N) of the number of candidate blocks for which the prediction error PE2 is to be obtained.

The processing procedure will be described below.

(1) The counter 21 is set to 0 and the prediction error PE1
Is TH, and the maximum value of the number of candidate blocks for obtaining the prediction error PE2 is M.

(2) The prediction error calculator 5 calculates L
The prediction error PE1 is obtained using 1 (<L) pixels, and the comparison unit 6 compares the prediction error PE1 with the threshold value TH.

(3) If the prediction error PE1 is equal to or less than the threshold value TH, the counter 1 is incremented, and the prediction error calculation section 5 calculates the prediction error PE2 using L2 (L1 <L2 ≦ L) pixels. The comparison unit 7 calculates the prediction error PE2 and the minimum value M
IN_PE is compared, and the prediction error PE2 is reduced to the minimum value MIN_PE.
If it is smaller than PE, the prediction error PE2 is newly set to the minimum value M
It is stored in the register 8 together with the motion vector as IN_PE.

(4) The processes from (2) to (3) are performed until the value of the counter 21 becomes equal to M or until there are no more candidate blocks. Finally, the motion vector stored in the register 8 is converted to the optimum motion vector. Vector.

The processing when the prediction error PE1 is larger than the threshold value TH in all the candidate blocks is the same as that in the first embodiment.

The amount of calculation required for block matching in the above method will be described. Here, sub indicates subtraction, add indicates addition, and ops indicates the amount of calculation. In addition, a rectangular block having a block size of 16 × 16, and a search range of a motion vector is −15 to +1 in both the horizontal direction and the vertical direction.
It is assumed that the accuracy of the five rectangular areas and the motion vectors is in units of one pixel.

Assuming that L1 = 16 and L2 = 256, the amount of calculation required in the above processing is the calculation of the prediction error PE1: 16 (sub) +15 (ad
d) = 31 ops Comparison of prediction error PE1 with threshold value TH: 1 ops Calculation of prediction error PE2: 240 (sub) +240 (a)
dd) = 480 ops Comparison between the prediction error PE2 and the minimum value MIN_PE: 1 ops. Therefore, if the number of candidate blocks for which the prediction error PE1 is obtained is β and the number of candidate blocks for which the prediction error PE2 is obtained is α, (31 + 1) × β + (480 + 1) × α =
32β + 481αops. However, the maximum value is 30
752 + 481 Mops.

From the above, the prediction error PE1 is set only for the candidate blocks whose prediction error PE1 is equal to or smaller than the threshold value TH.
By obtaining 2, a motion vector can be obtained with a smaller amount of calculation than in a full search.

FIG. 3 is a diagram showing an example of a circuit for realizing a third embodiment of the motion vector detecting method according to the present invention. A counter 31, a comparing unit 32, a threshold value TH, and an M updating unit 33 are added to the circuit of FIG. The counter 31 has a prediction error P
The number of candidate blocks in which E1 is equal to or smaller than the threshold value TH is counted. The comparing unit 32 compares the value of the counter 31 with the maximum value M (<N) of the number of candidate blocks for which the prediction error PE2 is to be obtained. When the value of the counter 31 is larger than M, the threshold TH / M updating unit 33 updates the threshold TH in a decreasing direction and updates M in an increasing direction.

The processing procedure will be described below.

(1) Counter 31 is set to 0, prediction error PE1
Is TH, and the maximum value of the number of candidate blocks for obtaining the prediction error PE2 is M.

(2) The prediction error calculator 5 calculates L
The prediction error PE1 is obtained using 1 (<L) pixels, and the comparison unit 6 compares the prediction error PE1 with the threshold value TH.

(3) If the prediction error PE1 is equal to or smaller than the threshold value TH, the prediction error calculation unit 5 obtains the prediction error PE2 using L2 (L1 <L2 ≦ L) pixels, and the comparison unit 7 calculates the prediction error PE2. PE2 is compared with the minimum value MIN_PE, and the prediction error P
If E2 is smaller than the minimum value MIN_PE, the prediction error P
E2 is newly stored in the register 8 together with the motion vector as the minimum value MIN_PE, and the counter 31 is incremented. The comparator 32 compares the value of the counter 31 with M. If the value of the counter 31 is larger than M, the threshold is exceeded. The value TH and M updating unit 33 updates the threshold value TH in the decreasing direction, and updates M in the increasing direction.

(4) The processes from (2) to (3) are performed on all the candidate blocks, and the motion vector finally stored in the register 8 is set as the optimum motion vector.

The processing when the prediction error PE1 is larger than the threshold value TH in all the candidate blocks is the same as that in the first embodiment.

The amount of calculation required for block matching in the above method will be described. Here, sub indicates subtraction, add indicates addition, and ops indicates the amount of calculation. In addition, a rectangular block having a block size of 16 × 16, and a search range of a motion vector is −15 to +1 in both the horizontal direction and the vertical direction.
It is assumed that the accuracy of the five rectangular areas and the motion vectors is in units of one pixel.

Assuming that L1 = 16 and L2 = 256, the amount of calculation required in the above processing is the calculation of the prediction error PE1: 16 (sub) +15 (ad
d) = 31 ops Comparison of prediction error PE1 with threshold value TH: 1 ops Calculation of prediction error PE2: 240 (sub) +240 (a)
dd) = 480 ops Comparison between the prediction error PE2 and the minimum value MIN_PE: 1 ops. Therefore, assuming that the number of candidate blocks for which the prediction error PE2 is obtained is α, (31 + 1) × 31 ² + (480+
1) × α = 30752 + 481αpos.

As described above, the prediction error PE1 is set only for the candidate block whose prediction error PE1 is equal to or smaller than the threshold value TH.
By obtaining 2, a motion vector can be obtained with a smaller amount of calculation than in a full search. It should be noted that it is also possible to determine the maximum value of M, and when the value of the counter 31 becomes equal to that value, terminate the search thereafter to reduce the amount of calculation to a certain value or less.

FIG. 4 is a diagram showing an example of a circuit for realizing a fourth embodiment of the motion vector detecting method according to the present invention. In the present embodiment, the threshold value TH is updated at a predetermined point in time in accordance with the ratio R of the number of candidate blocks for which the prediction error PE2 has been obtained to the number of candidate blocks for which the prediction error PE1 has been obtained. That is, if R is less than the predetermined value RTH, the threshold value TH is updated in the increasing direction, and if R is greater than RTH, the threshold value TH is updated in the decreasing direction. R is RTH
If it is equal to, do not update.

The circuits of FIG. 1 are provided with counters 41 and 42, a divider 43, a comparator 44, and a threshold TH updater 45. The counter 41 counts the number of candidate blocks for which the prediction error PE1 has been obtained. The counter 42 counts the number of candidate blocks for which the prediction error PE1 is equal to or less than the threshold value TH,
That is, the number of candidate blocks obtained up to the prediction error PE2 is counted. The division unit 43 obtains the ratio R of the candidate block for which the prediction error PE2 has been obtained to the number of candidate blocks for which the prediction error PE1 has been obtained. The comparison unit 44 has R and RTH
Compare with The threshold TH updating unit 45 updates TH in the increasing direction when R is less than RTH, and updates TH in the decreasing direction when R is greater than RTH. If R is equal to RTH, do not update.

The processing procedure will be described below.

(1) The counters 41 and 42 are set to 0, the threshold value of the prediction error PE1 is set to TH, and the threshold value of the ratio R is set to RTH.

(2) The prediction error calculator 5 calculates L
The prediction error PE1 is obtained using 1 (<L) pixels, the counter 41 is incremented, and the comparison unit 6 calculates the prediction error P1.
E1 is compared with a threshold value TH.

(3) If the prediction error PE1 is equal to or less than the threshold value TH, the counter 42 is incremented, and the prediction error calculation unit 5 obtains the prediction error PE2 using L2 (L1 <L2 ≦ L) pixels. The comparison unit 7 compares the prediction error PE2 with the minimum value MIN_PE, and determines that the prediction error PE2 is the minimum value MIN_PE.
If it is smaller than _PE, the prediction error PE2 is newly stored in the register 8 together with the motion vector as the minimum value MIN_PE.

(4) The ratio of the value of the counter 42 to the value of the counter 41 in the division unit 43, that is, the prediction error PE with respect to the number of candidate blocks for which the prediction error PE1 has been obtained so far.
The ratio R of the number of candidate blocks obtained up to 2 is obtained, and the comparing unit 44 compares R with RTH. If the ratio R is larger than RTH, the threshold TH updating unit 45 increases the threshold TH in the decreasing direction. When the ratio R is smaller than RTH, the threshold value TH is updated in the increasing direction.

(5) The processes from (2) to (4) are performed on all the candidate blocks, and the motion vector finally stored in the register 8 is set as the optimum motion vector.

Here, the processing of (4) can be performed for each candidate block, or each time the processing of a predetermined number of candidate blocks is completed. Further, the value of the RTH for comparison with the ratio R may have a range. That is, if α <RTH <β, the threshold value TH is updated in a decreasing direction when R> β, and the threshold value TH is updated when R <α.
It is also possible to update the threshold TH in a decreasing direction when R> α, and to update the threshold TH in an increasing direction when R <β. Furthermore, the ratio R does not necessarily need to be obtained by division,
For example, it is possible to use a ROM table that inputs the value of the counter 41 and the value of the counter 42 and outputs the ratio R according to a predetermined rule.

The processing when the prediction error PE1 is larger than the threshold value TH in all the candidate blocks is the same as that in the first embodiment.

The amount of calculation necessary for block matching in the above method will be described. Here, sub indicates subtraction, add indicates addition, and ops indicates the amount of calculation. In addition, a rectangular block having a block size of 16 × 16, and a search range of a motion vector is −15 to +1 in both the horizontal direction and the vertical direction.
It is assumed that the accuracy of the five rectangular areas and the motion vectors is in units of one pixel.

Assuming that L1 = 16 and L2 = 256, the amount of calculation required in the above processing is as follows: Calculation of prediction error PE1: 16 (sub) +15 (ad
d) = 31 ops Comparison of prediction error PE1 with threshold value TH: 1 ops Calculation of prediction error PE2: 240 (sub) +240 (a)
dd) = 480 ops Comparison between the prediction error PE2 and the minimum value MIN_PE: 1 ops. Therefore, assuming that the number of candidate blocks for which the prediction error PE2 is obtained is α, (31 + 1) × 31 ² + (480+
1) × α = 30752 + 481αops.

As described above, the prediction error PE1 is set only for the candidate blocks whose prediction error PE1 is equal to or smaller than the threshold value TH.
By obtaining 2, a motion vector can be obtained with a smaller amount of calculation than in a full search.

Note that the maximum value of the number of candidate blocks for which the prediction error PE2 is obtained is M (<N), and when the number of candidate blocks for which the prediction error PE2 is obtained becomes M, the search is terminated thereafter. It is also possible to reduce the amount of calculation below a certain value.

FIG. 5 is a diagram showing an example of a circuit for realizing a fifth embodiment of the motion vector detecting method according to the present invention. In this embodiment, the circuit of FIG.
3 are added.

The counter 51 has previously calculated the prediction error PE1.
Is counted below the threshold value TH.
The comparing unit 52 compares the value of the counter 51 with the value of the maximum value M (<N) of the number of candidate blocks for obtaining the prediction error PE2. The memory 53 is a memory for storing information (prediction error PE1, motion vector) of a candidate block in which the prediction error PE1 is equal to or less than the threshold value TH. Here, the prediction error P
It is assumed that information is stored using a binary tree based on the value of E1.

The processing procedure will be described below.

(1) Counter 51 is set to 0, prediction error PE1
Is TH, and the maximum value of the number of candidate blocks for obtaining the prediction error PE2 is M.

(2) The prediction error calculation unit 5 calculates a prediction error PE1 for all candidate blocks using L1 (<L) pixels in the block, and the comparison unit 6 calculates the prediction error PE1 and the threshold value TH. And the prediction error PE1 becomes equal to the threshold TH.
If not, the motion vector to the candidate block and the prediction error PE1 are stored in the memory 53 in the form of a binary tree, and the counter 51 is incremented.

(3) The value of the counter 51 and M
And if the value of the counter 51 is equal to or less than M, for all the candidate blocks stored in the memory 53,
The prediction error calculation section 5 calculates a prediction error PE2 using L2 (L1 <L2 ≦ L) pixels. When the value of the counter 51 is M
If the difference is larger than the prediction error PE1
, And the prediction error calculation unit 5 calculates a prediction error PE2 for the M candidate blocks by using L2 pixels. When calculating the prediction error PE2, the comparison unit 7 calculates the prediction error PE2 and the minimum value MIN. Compare PE and predict error PE
2 is the minimum value MIN If smaller than PE, prediction error PE
2 is newly added to the minimum value MIN It is stored in the register 8 together with the motion vector as PE.

(4) Finally, the motion vector stored in the register 8 is set as the optimum motion vector.

The processing in the case where the prediction error PE1 is larger than the threshold value TH in all the candidate blocks is the same as that in the first embodiment.

The amount of calculation required for block matching in the above method will be described. Here, sub indicates subtraction, add indicates addition, and ops indicates the amount of calculation. In addition, a rectangular block having a block size of 16 × 16, and a search range of a motion vector is −15 to +1 in both the horizontal direction and the vertical direction.
It is assumed that the accuracy of the five rectangular areas and the motion vectors is in units of one pixel.

Assuming that L1 = 16 and L2 = 256, the amount of calculation required in the above processing is: calculation of prediction error PE1: 16 (sub) +15 (ad
d) = 31 ops Comparison of prediction error PE1 with threshold value TH: 1 ops Calculation of prediction error PE2: 240 (sub) +240 (a)
dd) = 480 ops prediction error PE2 and minimum value MIN The comparison of PEs is 1 ops, and if α is the number of candidate blocks for which the prediction error PE2 is obtained, (31 + 1) × 31 ² + (480+
1) × α = 30752 + 481αops.

The maximum value is 30752 + 481 Mops.

From the above, by calculating the prediction error PE2 only for the M or less candidate blocks whose prediction error PE1 is equal to or less than the threshold value TH, the calculation amount is smaller than that of the full search, and is reduced to a certain value or less. Can be reduced.

[0101]

According to the present invention, a motion vector detecting method for detecting a motion vector by performing block matching between a prediction target image composed of L pixels and a reference image in video coding using motion compensated prediction. And L1
A first step of calculating a first prediction error for (L1 <L) pixels by block matching between the candidate block and the image to be predicted;
A second step of calculating a second prediction error by block matching between a candidate block and a prediction target image for L2 (L1 <L2 ≦ L) pixels only in the following cases:
If the second prediction error obtained is the smallest second prediction error, line the motion vector and a third step of the motion vector of the prediction target image, each candidate block
Steps 1 to 3 are repeated for all candidate blocks.
Since the motion vector is repeatedly obtained until the block matching for the block is completed, the amount of calculation and memory required for the block matching can be significantly reduced, and a highly accurate motion vector can be detected.

[Brief description of the drawings]

The first embodiment of a motion vector detecting method of the present invention; FIG
FIG. 3 is a diagram illustrating an example of a circuit to be realized .

The second embodiment of the motion vector detecting method of the present invention; FIG
FIG. 3 is a diagram illustrating an example of a circuit to be realized .

The third embodiment of the motion vector detecting method of the present invention; FIG
FIG. 3 is a diagram illustrating an example of a circuit to be realized .

The fourth embodiment of the motion vector detecting method of the present invention; FIG
FIG. 3 is a diagram illustrating an example of a circuit to be realized .

[5] The fifth embodiment of the motion vector detecting method of the present invention
FIG. 3 is a diagram illustrating an example of a circuit to be realized .

FIG. 6 is a diagram illustrating pixels used when obtaining a prediction error for L1 pixels.

FIG. 7 is a diagram illustrating a means for obtaining a minimum value of a prediction error PE1.

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int.Cl. ⁷ , DB name) H04N ^7/ 24-7/68

Claims (4)

(57) [Claims] 1. A motion vector detecting method for detecting a motion vector by performing block matching between a prediction target image composed of L pixels and a reference image when a moving image is encoded using motion compensated prediction. A first step of calculating a first prediction error by block matching between a candidate block and a prediction target image for (L1 <L) pixels, and only when the first prediction error is equal to or less than a predetermined threshold T , L2
A second step of calculating a second prediction error for the (L1 <L2 ≦ L) pixels by block matching between the candidate block and the image to be predicted, and a second step in which the obtained second prediction error is the minimum. if the prediction error, the line the motion vector and a third step of the motion vector of the prediction target image, each candidate block
Then, the first to third steps are performed on all candidate blocks.
Against repeated until block matching has been completed, the motion vector detecting method characterized by obtaining a motion vector. 2. A motion vector detecting method for detecting a motion vector by performing block matching between a prediction target image composed of L pixels and a reference image during video encoding using motion compensated prediction, comprising: A first step of calculating a first prediction error by block matching between a candidate block and a prediction target image for (L1 <L) pixels, and only when the first prediction error is equal to or less than a predetermined threshold T , L2
A second step of calculating a second prediction error for the (L1 <L2 ≦ L) pixels by block matching between the candidate block and the image to be predicted, and a second step in which the obtained second prediction error is the minimum. if the prediction error, the line the motion vector and a third step of the motion vector of the Yoroku target image, each candidate block
There, the third step from the first, or second step is executed a predetermined number of times, repeatedly until the block <br/> click matching against all of the candidate blocks is terminated, characterized in that a motion vector is obtained A motion vector detection method. 3. A motion vector detecting method for detecting a motion vector by performing block matching between a prediction target image composed of L pixels and a reference image in video coding using motion compensated prediction. A first step of calculating a first prediction error by block matching between a candidate block and a prediction target image for (L1 <L) pixels, and only when the first prediction error is equal to or less than a predetermined threshold T , L2
A second step of calculating a second prediction error by block matching between the candidate block and the prediction target image for (L1 <L2 ≦ L) pixels, and when the second step is performed a predetermined number of times, A third step of updating the threshold value T to be decreased; and a fourth step of setting the motion vector as the motion vector of the prediction target image if the obtained second prediction error is the minimum second prediction error. and a step, lines each candidate block
There, a fourth step from the first, to all of the candidate blocks
Against repeated until block matching has been completed, the motion vector detecting method characterized by obtaining a motion vector. 4. A motion vector detecting method for detecting a motion vector by performing block matching between a prediction target image composed of L pixels and a reference image when a moving image is encoded using motion compensated prediction. A first step of calculating a first prediction error by block matching between a candidate block and a prediction target image for (L1 <L) pixels, and only when the first prediction error is equal to or less than a predetermined threshold T , L2
A second step of calculating a second prediction error by block matching between the candidate block and the prediction target image for (L1 <L2 ≦ L) pixels, and a second prediction error with respect to the number of times the first prediction error is obtained A third step of obtaining a ratio R of the number of times of obtaining the above, if the ratio R is larger than a predetermined ratio RT, the threshold T is updated so as to decrease , and the ratio R is smaller than the predetermined ratio RT. In this case, a fourth step of updating the threshold value T to be increased, and if the obtained second prediction error is the minimum second prediction error, the motion vector is replaced with the motion vector of the prediction target image. Is performed for each candidate block, and the first to fifth steps are performed on all the candidate blocks.
Against repeated until block matching has been completed, the motion vector detecting method characterized by obtaining a motion vector.