JPH08168054A - Moving vector detecting device - Google Patents

Abstract

PURPOSE: To obtain continuity of a moving vector nearby a boundary part by regarding the moving vector already detected as an initial candidate vector as to adjacent divided images. CONSTITUTION: An initial deviation vector selecting circuit 25 inputs detected vectors SV and SVL and/or SVR of blocks at the peripheries of >2 current blocks to be detected outputted from an in-divided-image detected vector memory 21, and a left and/or right vector memory 22 or 23 as initial candidate vectors SF. The circuit 25 moves the last field signal SE of a divided image by as much as the vector SF to perform block matching with the current field signal SD of an enlarged divided image, and an initial candidate vector having the shortest matching distance is supplied as an initial deviation vector Vo to a moving vector detecting circuit body 26. The moving vector of the current block to be detected is detected on the basis of the signals SE, SD, and Vo and outputted.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a motion vector detecting device, and is particularly suitable for application to a device for dividing an image and obtaining motion vectors in a plurality of divided images in parallel.

[0002]

2. Description of the Related Art A motion vector indicates the magnitude and direction of motion of a moving object in an image.
It is used for interframe coding in high-efficiency coding of video signals, field interpolation in field number conversion in television system conversion, and the like.

As a method of detecting a motion vector, a pattern matching method is generally used to detect a motion vector by using the similarity of signal patterns between frames (Japanese Patent Laid-Open No. 55-162).
No. 683, JP-A-55-162684), and generally an iterative gradient method (Japanese Patent Laid-Open Publication No. Sho-62-163) that estimates the amount of motion by physically corresponding the intra-frame signal gradient and the inter-frame signal difference value. No. 60-158786) is known.

Some of the iterative gradient methods include an initial displacement vector (Japanese Patent Laid-Open No. Sho 62-62) in order to improve the motion vector detection accuracy.
-206980). Also, in the pattern matching method, there is a method that uses the initial displacement vector so that the target area to be subjected to the matching processing becomes narrow.

In order to execute the motion vector detection at high speed, for example, as shown in FIG. 2, a plurality of partial images PA (PB, PC, PD) of one image (four in FIG. 2) are provided.
It has already been proposed that the motion vector is divided into blocks and the motion vector is obtained in parallel in block units in the divided images PA, PB, PC and PD. In this way, a plurality of divided images PA, P
When detecting motion vectors in parallel in B, PC, and PD, in order to accurately detect the motion vector of a block at the boundary between a certain divided image and an adjacent divided image,
As shown in FIG. 2, the divided image is connected to the overlapping image to enlarge the divided image, and motion vector detection is performed for all of the enlarged divided image areas. Then, the motion vector of the divided image area is output excluding the motion vector of the overlapping image area.

When the divided images are enlarged by connecting the overlapped images, the detection accuracy is higher than that in the case where the search region at the corresponding position of the block to be detected is enlarged and not enlarged, and the blocks of the original divided image region are As the initial displacement vector required at the time of detection, the motion vector obtained for the overlap image area can be used as a candidate, and the detection accuracy can be improved.

[0007]

However, the motion vector is detected for the enlarged divided image enlarged by connecting the overlapped image to the divided image, and the divided image is cut out from the enlarged divided image to detect the movement of the plurality of divided images. When the vectors are connected, even if the overlapped image area is provided, the motion vector is detected for each enlarged divided image independently of the other enlarged divided images. However, there was a drawback that it became an unnatural connection (discontinuity).
As a result, there is also a drawback that the image processed using the motion vector and the reproduced image are distorted.

Therefore, even when one image is divided into a plurality of parts and motion vectors are obtained in parallel for a plurality of divided images, the motion is performed so as to have the same degree of continuity as the motion vector detected without dividing the image. It is desired to detect the vector.

[0009]

In order to solve such a problem, according to the present invention, a divided image motion vector detection unit for each divided image obtained by dividing one image is provided, and each divided image motion vector detection unit is , A motion vector detection device for detecting a motion vector between video signals separated by at least one field for each block of m lines × n pixels (m and n are integers), wherein each divided image motion vector detection unit includes: A motion that employs a motion vector detection method that involves a process in which an already-detected motion vector of a block around a block to be detected is used as an initial candidate vector, and one of the plurality of initial candidate vectors is selected as an initial displacement vector. The vector detecting device is characterized in that it is configured as follows.

That is, each of the divided image motion vector detection units uses the motion vector already detected by each of the other divided image motion vector detection units corresponding to the divided images adjacent to the divided image of its own as the position of the block to be detected. Some of them are characterized by having an adjacent divided image already-detected vector holding / selecting unit which is used as an initial candidate vector.

[0011]

According to the present invention, there is provided a divided image motion vector detection unit for each divided image obtained by dividing one image into a plurality of divided images, and each divided image motion vector detection unit has at least one field interval for each block of m lines × n pixels. A method for detecting a motion vector between video signals distant from each other, further, each divided image motion vector detection unit, the already detected motion vector of the surrounding block of the block to be detected as an initial candidate vector,
It is premised on a motion vector detection device that employs a motion vector detection method that involves the process of selecting one from a plurality of initial candidate vectors and setting it as an initial displacement vector.

Conventionally, each divided image motion vector detection unit has a closed configuration in which signals are not exchanged with each other, so that it is possible to use information on motion vectors already detected by other divided image motion vector detection units. Therefore, discontinuity of the motion vector occurs at the division boundary.

Therefore, in the present invention, each of the divided image motion vector detection units receives the motion vector already detected by each of the other divided image motion vector detection units corresponding to the divided images adjacent to the divided image relating to itself. Depending on the position of the block to be detected, an adjoining divided image already detected vector holding / selecting means for setting the initial candidate vector is provided, and when a block near the boundary becomes the block to be detected, the motion already detected for the adjoining divided images is detected. The vector can be used as the initial candidate vector, and the continuity of the motion vector near the boundary is achieved.

[0014]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of a motion vector detecting device according to the present invention will be described below in detail with reference to the drawings.

Here, FIG. 1 is a block diagram showing the configuration of each divided image motion vector detection unit in the motion vector detection device of this embodiment, and FIG. 3 is the overall configuration of the motion vector detection device of this embodiment. It is a block diagram showing.

For simplification of description, the motion vector detecting device of this embodiment divides one image into four equal parts in the horizontal direction by the vertical dividing line (see FIG. 2) and four parts. It is assumed that the motion vector of the image is detected in parallel.
Further, the motion vector detecting device of this embodiment uses the previous field signal as a reference and the current field signal as the search side signal at the corresponding position, and the size of the detection block is 8 pixels × 8 lines. And

In FIG. 3, the motion vector detecting apparatus 10 of this embodiment delays the input digital video signal (which may be any of a luminance signal, a color signal, a composite video signal, etc.) by one field. A field delay circuit 11 is provided so that the current field signal and the previous field signal synchronized with the one field delay circuit 11 can be taken out. The extracted current field signal and previous field signal are respectively applied to the corresponding image division unit 12 or 13 of the memory configuration, for example.

In this embodiment, the current field image division unit 12 to which the signal of the current field on the search side is input.
Converts the current field signal into a current field signal of a divided image obtained by dividing one image into four, and (the position of the divided image)
A divided image motion vector detection unit 14P provided for each
A, 14PB, 14PC, 14PD. In addition, the current field image division unit 12 obtains a right-side overlap image signal and a left-side overlap image signal for each of the divided images PA, PB, PC, PD (see FIG. 2) from the current field signal and responds. The divided image motion vector detecting units 14PA, 14PB, 14PC, and 14PD are provided. A meaningless left-side overlap image signal is given to the motion vector detection unit 14PA related to the leftmost divided image PA, and a meaningless right-side overlap is applied to the motion vector detection unit 14PD related to the rightmost divided image PD. An image signal is given.

In the present embodiment, the front field image dividing section 13 to which the signal of the front field on the reference side is input.
Converts the previous field signal into a previous field signal of a divided image obtained by dividing one image into four, and (the position of the divided image)
A divided image motion vector detection unit 14P provided for each
A, 14PB, 14PC, 14PD. The image division unit 13 for the previous field is the image division unit 1 for the current field.
The division processing is performed in synchronization with 2. In the motion vector detecting device 10 of this embodiment, unlike the conventional one,
On the reference side, the overlapping image area is not used for detecting the motion vector.

At the same time, the current field image dividing unit 12 and the previous field image dividing unit 13 are arranged so that the positions of blocks to be detected in the divided images PA, PB, PC and PD are different in the vertical direction. In addition, the divided image motion vector detection units 14PA, 14PB, 14PC, 14PD
Signal is output to. For example, when the block at the k-th position in the vertical direction of the divided image PA is the detection target, such as the block indicated by the upper right hatch in FIG. 4, the block at the k−1-th position is detected in the divided image PB. The target field image partitioning unit 12 and the current field image partitioning unit 12 are configured such that the block at the k-2 position is detected as the target of the divided image PC and the block at the k-3 position is targeted as the target of the divided image PD. The front field image division unit 13 adjusts the output timing.

For example, each of the divided images PA, PB,
When the leftmost block of the PC or PD (the block with the upper right hatch) is the detection target, the divided image is different, but the block to the left of it (the block with the upper left hatch)
This is because the motion vector of is already detected.

Each divided image motion vector detection unit 14PA,
The 14PB, 14PC, and 14PD have the configuration shown in FIG. 1 described later and are the target divided images PA, PB, and P.
For C and PD, a motion vector is detected for each block of 8 pixels × 8 lines, and the detected motion vector is output to the motion vector output unit 15. Motion vector output unit 15
Is mainly composed of a memory, for example, and outputs the motion vector of each divided image PA, PB, PC, PD in a predetermined order determined for one image.

It should be noted that each divided image motion vector detection unit 14
The motion vectors output from the PAs, 14PBs, 14PCs, and 14PDs are supplied to not only the motion vector output unit 15 but also other divided image motion vector detection units related to adjacent divided images, which will be described later. In addition, the divided image motion vector detecting unit appropriately uses it for detecting a motion vector.

Next, the divided image motion vector detection unit 14
The detailed configuration and operation of (14PA, 14PB, 14PC, 14PD) will be described with reference to FIG.

In FIG. 1, the divided image motion vector detection unit 14 includes a divided image enlarging circuit 20, an already detected vector memory in a divided image 21, a left vector memory 22, a right vector memory 23, an initial candidate vector switching circuit 24, an initial deviation vector. It is composed of a position vector selection circuit 25 and a motion vector detection circuit body 26.

The current field signal SA of the divided image, the current field signal SB of the left overlapped image and the current field signal SC of the right overlapped image output from the above-mentioned image division unit for current field 12 are divided image expansion circuit 20. Given to. The divided image enlarging circuit 20 forms the current field signal SD relating to the enlarged divided image (see FIG. 2) based on these signals SA, SB and SC to form the initial displacement vector selection circuit 25 and the motion vector detection circuit body 26. Give to.

On the other hand, the front field signal SE of the divided images output from the front field image division unit 13 is output.
Is applied to the initial displacement vector selection circuit 25 and the motion vector detection circuit body 26 without being enlarged.

The current field image dividing section 12 may be configured to output the current feed signal of the enlarged divided image with the left and right overlapping images from the beginning. In this case, the divided image is generated. The expansion circuit 20 is omitted.

The divided image motion vector detection unit 14 uses an already detected vector memory 21 in the divided image as an already detected vector memory for storing the already detected motion vector.
It has three types of memories, a left vector memory 22 and a right vector memory 23.

The already detected vector memory 21 in the divided image is
The divided image motion vector detection unit 14 stores the motion vector V already detected, and outputs the already detected vector SV of the surrounding blocks of the current block to be detected.

The left vector memory 22 stores the right end block of the divided image which has already been detected by the divided image motion vector detection unit related to the divided image adjacent to the left of the divided image targeted by the divided image motion vector detection unit 14. The motion vector (left vector) VL is stored, and when the current detection target block is the leftmost block of the divided image,
The already-detected vector SVL of the surrounding block (the block on the left side or the block on the upper left side) is output. For example, the divided image motion vector detection unit 14 may be the detection unit 1
If it is 4B (see FIG. 3), the left vector VL already detected in the divided image motion vector detection unit 14A is stored in the left vector memory 22.

On the other hand, the right vector memory 23 stores the left end of the divided image which has already been detected by the divided image motion vector detection unit related to the divided image on the right of the divided image targeted by the divided image motion vector detection unit 14. The motion vector (right vector) VR of the block is stored, and when the current detection target block is the right end block of the divided image, the already detected vector SVR of the surrounding blocks (upper right block and the like) is output. . For example, if the divided image motion vector detection unit 14 is the detection unit 14B (see FIG. 3), the right vector VR already detected by the divided image motion vector detection unit 14C is stored in the right vector memory 23.

The initial candidate vector switching circuit 24 includes an already detected vector memory 21 in the divided image and a left vector memory 2
2 and / or the detected vector SV, SVL and / or SVR of two or more surrounding blocks of the current block to be detected, which are output from the right and left vector memory 23, are selected as candidates for the initial displacement vector (initial candidate vector). The SF is sequentially input to the initial displacement vector selection circuit 25.

For example, if the block to be detected is the block located at the left end of the divided image (the second and subsequent ones in the vertical direction), the already-detected vector memory 21 in the divided image is detected.
From, the motion vectors for the blocks immediately above and to the upper right are selected, and from the left vector memory 22, the motion vectors for the blocks to the left and the upper left are selected. Further, for example, if the block to be detected is the block located at the right end of the divided image (the second and subsequent ones in the vertical direction), the detected vector memory 21 in the divided image is located on the left side, directly above, and above. The motion vector for the upper left block is selected, and the motion vector for the upper right block is selected from the right vector memory 23. Further, for example, if the block to be detected is a block in the central portion that is not located at the left end or right end of the divided image (the second and subsequent ones in the vertical direction), the detected image vector inside the divided image 21 is moved to the upper right corner. , Motion vectors for blocks immediately above, to the upper left and to the left.

The initial displacement vector selection circuit 25 moves the previous field signal SE of the divided image by the initial candidate vector SF to perform block matching with the current field signal SD of the enlarged divided image, and the initial candidate having the shortest matching distance. A vector is found, the vector is determined as the initial displacement vector Vo, and it is given to the motion vector detection circuit body 26.

The motion vector detection circuit body 26 detects the motion vector V for the current block to be detected using the previous field signal SE of the divided image, the current field signal SD of the enlarged divided image and the initial displacement vector Vo. And output. The motion vector V is fed back to the in-divided-image already-detected vector memory 21 inside the divided image motion vector detection unit 14, and to the outside of the divided image motion vector detection unit 14, the motion vector output unit. 15 and other divided image motion vector detection units related to adjacent divided images.

The motion vector detecting method adopted by the motion vector detecting circuit main body 26 may be any detecting method as long as it processes in block units using the initial displacement vector Vo. Iterative gradient method and pattern matching method can be applied.

FIGS. 5 and 6 are explanatory diagrams of the motion vector detection circuit according to the iterative calculation method in which the iterative gradient calculation is executed twice.

In FIG. 5, the initial displacement vector selection circuit 41 selects the previous field signal SE and the current field signal SD from the detected motion vectors in the vicinity of the detected block stored in the detected vector memory 40.
Initial displacement vector Vo based on block matching of
Is selected. A vector Vp of the deviation from the initial displacement vector Vo is obtained by the gradient method calculation by the first gradient method calculation unit 42 using the initial displacement vector Vo, the previous field signal SE and the current field signal SD, and the addition is performed. This vector is corrected by the device 43 (see FIG. 6). Further, the vector Vo + Vp, the previous field signal S
The vector Vo + Vp is calculated by the gradient method calculation by the second gradient method calculation unit 44 using E and the current field signal SD.
A vector Vq corresponding to the deviation from is calculated, and this vector is corrected by the adder 45 (see FIG. 6) and output as the final motion vector V (= Vo + Vp + Vq) and is also given to the already detected vector memory 40. Stored.

Therefore, the divided image motion vector detection unit 14
5 adopts the iterative gradient method, the constituent parts of the already-detected vector memory 21 in the divided image, the left vector memory 22, the right vector memory 23, and the initial candidate vector switching circuit 24 in FIG. The detected vector memory 40 corresponds, and the initial displacement vector selection circuit 25 in FIG. 1 corresponds to the initial displacement vector selection circuit 41 in FIG. In addition, the first gradient method computing unit 42, the adder 43, and the second gradient method computing unit 4 in FIG.
4 and the adder 45 constitute the motion vector detection circuit main body 26 in FIG.

In the above configuration, the input video signal of one image is input to the one-field delay circuit 11, and the current field signal and the previous field signal which are synchronized before and after the one-field delay circuit 11 are taken out and correspond to each other. The divided image motion vector detection units 14PA, 14PB, 14PC, 1 are input to the image division units 12 and 13 and are divided into images, and the divided image motion vector detection units 14PA, 14PB, 14PC, and
It is input to each 4PD.

Each divided image motion vector detection unit 14 (14
PA, 14PB, 14PC, 14PD), for the current field signal, the divided image enlarging circuit 20 enlarges the divided image, and the current field signal SD related to the enlarged divided image and the previous field signal SE related to the divided image are initially biased. It is input to the position vector selection circuit 25 and the motion vector detection circuit body 26.

The initial displacement vector selection circuit 25 is supplied with the current target vector output from the already detected vector memory 21 in the divided image, the left vector memory 22 and / or the right vector memory 23 via the initial candidate vector switching circuit 24. The already detected vectors SV, SVL, and / or SVR of two or more surrounding blocks of the detection target block are sequentially input as the initial candidate vector SF. Initial deviation vector selection circuit 2
5 performs block matching with the current field signal SD of the enlarged divided image by moving the previous field signal SE of the divided image by the initial candidate vector SF, finds the initial candidate vector with the shortest matching distance, and sets it as the initial deviation vector. The position vector Vo is determined and given to the motion vector detection circuit body 26. As a result, the motion vector detection circuit main body 26 detects the motion vector V for the current block to be detected based on the previous field signal SE of the divided image, the current field signal SD of the enlarged divided image, and the initial displacement vector Vo. Is output.

Each divided image motion vector detection unit 14PA,
The motion vector V output from 14PB, 14PC, and 14PD is input to the motion vector output unit 15 and output in a predetermined order. Further, the motion vector V output from each of the divided image motion vector detection units 14PA, 14PB, 14PC, 14PD is fed back to the in-divided image already detected vector memory 21 inside the self detection unit, and the divided image motion vector It is output to the other divided image motion vector detection unit to the outside of the detection unit, and as a result, it is stored in the divided image already detected vector memory 21, the left vector memory 22 and / or the right vector memory 23 as a detected vector. , Is used to detect the motion vector of another block.

Therefore, according to the above-described embodiment, when each divided image motion vector detection unit detects the block in the vicinity of the boundary as the detection target block, the divided image motion vector detection units for the adjacent divided images have already detected. Since the initial deviation vector is also determined as the initial candidate vector using the motion vector as described above, the accuracy of determination of the initial deviation vector is improved, and as a result, the continuity of the motion vector is compensated even at the boundary of the divided images, The motion vector can be detected (with continuity) with the same accuracy as when the motion vector is detected without dividing the image.

Thus, in the image processing using the motion vector by the motion vector detecting device of this embodiment,
It is possible to prevent image distortion from occurring in the vicinity of the boundary of the divided images.

In the above embodiment, the motion vector is detected from the video signal different from one field, but the present invention is also applied to the motion vector detected from the video signal different from one frame (two fields). The present invention can also be applied to the case where a motion vector is detected from video signals different in three fields or more.

In the above embodiment, the motion vector is detected with the past video signal (previous field signal) as the reference side and the current video signal (current field signal) as the search side. The present invention can be applied to a motion vector detecting device in which the search side and the search side are opposite.

Further, in the above embodiment, one image is divided into four in the horizontal direction and the detection processing is shown. However, the number of divisions is not limited to this, and the division direction may be the vertical direction. Alternatively, a combination of vertical and horizontal division may be used. In consideration of a general raster scan (horizontal direction raster scan), horizontal direction division is preferable as in the above embodiment.

Furthermore, in the above embodiment, the size of the motion block detection block is 8 lines × 8 pixels, but this size may be any size. The overlapping image area may be determined in consideration of the above.

In the above embodiment, the detected vector for the adjacent divided image of the divided images related to a certain divided image motion vector detecting unit is stored in the left vector memory 22 and the right vector memory 23. A fixed delay line may be used instead of the vector memory, and these vector memories may be omitted and the initial candidate vector switching circuit 24 may directly access the already detected vector memory in the divided image in another divided image motion vector detection unit. You may do it. The term “adjacent divided image already detected vector holding and selecting means” in the claims includes these modified examples.

In the above embodiment, the motion vector of the already detected periodic block related to the same field as the block to be detected is used as the initial candidate vector. However, for the surrounding blocks not detected in the same field, The motion vector detected one field before may be used as the initial candidate vector.

[0053]

As described above, according to the present invention, the divided image motion vector detecting unit for each divided image obtained by dividing one image into a plurality of divided images corresponds to the divided image adjacent to the divided image relating to itself. Since a motion vector already detected by each of the divided image motion vector detection units of the above is provided with an adjacent divided image already detected vector holding selection means for setting it as an initial candidate vector depending on the position of the block to be detected, it is divided by the conventional device. It is possible to remove the discontinuity of the motion vector that has occurred in the vicinity of the boundary portion of the image, and it is possible to execute the vector detection equivalent to that of the motion vector detection device that does not divide the image.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a divided image motion vector detection unit according to an embodiment.

FIG. 2 is an explanatory diagram of a method of dividing one image and a region used for detection.

FIG. 3 is a block diagram showing an overall configuration of a motion vector detection device according to an embodiment.

FIG. 4 is an explanatory diagram showing that the vertical position of a detection target block is different in each divided image motion vector detection unit of the embodiment.

FIG. 5 is an explanatory diagram of a detection configuration according to the iterative gradient method.

FIG. 6 is an explanatory diagram showing correction from an initial deviation vector in the iterative gradient method.

[Explanation of symbols]

10 ... Motion vector detecting device, 11 ... 1 field delay circuit, 12 ... Current field image dividing unit, 13 ... Previous field image dividing unit, 14PA, 14PB, 14PC, 14PD ... Divided image motion Vector detection unit, 15 ... Motion vector output unit, 20 ... Divided image enlarging circuit, 21 ... Divided image already detected vector memory, 22 ... Left vector memory, 23 ... Right vector memory, 24 ... Initial candidate Vector switching circuit, 25 ... Initial displacement vector selection circuit, 26 ... Motion vector detection circuit body.

Claims (2)

[Claims] 1. A divided image motion vector detection unit for each divided image obtained by dividing one image into a plurality of images, wherein each divided image motion vector detection unit is a block of m lines × n pixels (m and n are integers). A motion vector detecting device for detecting a motion vector between video signals separated by at least one field for each, wherein each of the divided image motion vector detecting units initializes an already-detected motion vector of a block surrounding a block to be detected. In a motion vector detection device adopting a motion vector detection method that involves selecting one of a plurality of initial candidate vectors as a candidate vector and setting it as an initial displacement vector, , A motion vector already detected by each of the above-mentioned divided image motion vector detection units corresponding to a divided image adjacent to the divided image relating to itself is detected. Depending on the position of the target block, a motion vector detecting apparatus characterized by having adjacent divided images already detected vector holding selecting means for the initial candidate vector. 2. One image is divided into a plurality of pieces in the horizontal direction, and at the same point in time, the divided image motion vector detection unit for the divided image on the left side has a lower position in the vertical direction of the block to be detected. The motion vector detection device according to claim 1.