KR101895822B1 - Apparatus and method for profile based motion estimation - Google Patents

Abstract

The present invention relates to a profile information-based motion estimation apparatus and method, and more particularly, to a horizontal profile image generation unit that receives a reference image and a current image and generates a horizontal profile image of the reference image and a horizontal profile image of the current image. A vertical profile image generation unit receiving the reference image and the current image and generating a vertical profile image of the reference image and a vertical profile image of the current image; A horizontal profile image and a vertical profile image of the reference image are received and a reference block profile of a block in the reference image is generated and a horizontal profile image and a vertical profile image of the current image are input to generate a current block profile of a current intra- A block profile calculation unit for calculating a block profile; And a motion vector estimator for calculating a difference between the reference block profile and the current block profile to predict a coordinate difference of a block having the smallest difference between the profiles as a motion vector.

Description

[0001] Apparatus and method for profile-based motion estimation [

The present invention relates to a profile information-based motion estimation apparatus and method, and more particularly, to a profile information-based motion estimation apparatus and method for generating a profile image for a reference image and a current image when estimating a motion vector for block matching, And more particularly, to a profile information-based motion estimation apparatus and method for reducing a computation amount for motion prediction and enabling high-speed encoding by comparing only profile information with respect to a candidate motion vector of a block so as to predict a motion vector.

A key element that supports multimedia is image compression technology along with digitalization. In recent years, digitization of information is rapidly developing, and image compression technology is becoming very important.

In general, a digital video device such as a PC, an HDTV, or a mobile device converts an image signal compressed by an image compression method such as MPEG (Moving Picture Experts Group) -2, MPEG-4, or H. 264 into an MPEG-based image codec And displays it by encoding.

This video CODEC eliminates temporal redundancy based on the block matching method. In the block matching method, the image is divided into small blocks, and a portion similar to a block of the current encoded / decoded image is temporally compressed in a method of finding and utilizing a previous frame or a previously encoded / decoded image and transmitting the difference value .

To do this, it is necessary to find the most similar block between the previous frame and the current frame. This is referred to as Motion Estimation (ME), and a motion vector (MV) indicating a displacement, .

The concept of intermediate image frame Fi synthesis using the block matching method is shown in FIG.

1, in order to synthesize an intermediate image frame Fi between a previous frame Fn-1 and a current frame Fn, a motion vector MV between a block in a previous frame Fn-1 and a block Fn in the current frame Motion estimation is performed to synthesize blocks of the intermediate image frame Fi.

In motion estimation for video signal compression, the current image is divided into macroblocks, and error values such as a sum of absolute difference (SAD) and a sum of square difference (SSD) The motion vector is used as a motion vector.

In the case of H.264, a motion vector with the lowest cost is used for motion estimation with a cost considering the amount of data required to transmit the motion vector together with the SAD or SSD.

Here, in order to find a motion vector, it is necessary to calculate SAD or SSD for all the motion search areas, and find a position having the minimum value among them. For example, if the block size is N x N and the motion search area is 32x32, 1024 * (2NxN-1) addition / subtraction operations are required. If the image size is about 100Nx80N, 8192000 * (2NxN- . This causes a problem of causing a system load and delaying the processing speed due to a large amount of computation to occupy an operation up to 50% in the image encoding.

Korean Patent Publication No. 2004-0049214 (2004.06.11), Title of the Invention: High-speed motion vector estimation apparatus and method

SUMMARY OF THE INVENTION The present invention has been conceived to solve the problems described above, and it is an object of the present invention to provide a method and apparatus for generating a profile image for a reference image and a current image, The present invention also provides a profile information-based motion estimation apparatus and method for estimating a motion vector by comparing only profile information with respect to all pixels, thereby reducing the amount of calculation for motion prediction and enabling high-speed image encoding.

According to another aspect of the present invention, there is provided a profile information-based motion estimation method comprising the steps of: (a) receiving a reference image, which is a previous view image, from a reference image buffer, Receiving a current image, which is a current time image, from the current image; (b) generating a horizontal profile of the reference image and a horizontal profile of the current image by summing pixel values on the same horizontal line for each horizontal line of the reference image and the current image, respectively; (c) generating a vertical profile of a reference image and a vertical profile of a current image by summing pixel values on the same vertical line for each vertical line of the reference image and the current image, respectively; (d) calculating a horizontal profile (Href (i)) and a vertical profile (Vref (i)) of the block in the reference image using the horizontal profile and the vertical profile of the reference image, Calculating a horizontal profile (Hcur (i)) and a vertical profile (Vcur (i)) for the block in the current image using the horizontal profile and the vertical profile; (i) a vertical profile (Href (i)) and a horizontal profile (Hcur (i)) for a block in the current image, and a vertical profile (I)) and the horizontal profile (Hcur (i)) for the block of the current image, using the horizontal profile (Href (i)) and the vertical profile (Vref And a vertical profile (Vcur (i)), respectively; (i)) and a horizontal profile (Hcur (i)) of the block of the current image calculated by the motion vector estimation unit and a horizontal profile (Href (i)) and a vertical profile And estimating a coordinate difference of blocks having a minimum difference value between the profiles (Vcur (i)) as a motion vector.

Here, the step (d) may include calculating a horizontal profile Href (i) and a vertical profile Vref (i) of a block of the reference image, a horizontal profile Hcur (i) And calculating the profile (Vcur (i)) using the following equation.

[Mathematical Expression]

Href (x) = Href (x + X, y + Y + i)

Vref (i) = Vref (x + X + i, y + Y) - Href (x + X + i, y +

Hcur (i) = Hcur (x, y + i) - Hcur (x-N, y + i)

Vcur (i) = Vcur (x + i, y) - Hcur (x + i,

(X, Y) is the motion vector, and N is the pixel size in the horizontal and vertical directions of the block)

(I)) and a horizontal profile (Hcur (i)) for a block of the current image and a vertical profile (Hcur (i)) for a block of the reference image. Vcur (i)) as a Sum of Absolute Difference (SAD) or Sum of Square Difference (SSD) value using the following equation.

[Mathematical Expression]

SAD = {? | Href (i) - Hcur (i) | +? | Vref (i) - Vcur (i) |} x?

SSD = {? | Href (i) - Hcur (i) | ² +? Vref (i) - Vcur (i) | ² } x?

(Where N is a scale constant and N is a natural number)

According to an aspect of the present invention, there is provided a profile information-based motion estimation apparatus including a reference image buffer for outputting a reference image signal, A current image buffer for outputting a current image signal which is a current time image; A horizontal profile that receives the reference image and the current image and generates a horizontal profile of the reference image and a horizontal profile of the current image by summing the pixel values on the same horizontal line for each horizontal line of the reference image and the current image, An image generation unit; A vertical profile that receives the reference image and the current image and generates a vertical profile of the reference image and a vertical profile of the current image by summing pixel values on the same vertical line for each vertical line of the reference image and the current image, An image generation unit; (Href (i)) and a vertical profile (Vref (i)) for a block in the reference image using the horizontal profile and the vertical profile of the reference image, and using the horizontal profile and the vertical profile of the current image A block profile calculation unit for calculating a horizontal profile (Hcur (i)) and a vertical profile (Vcur (i)) for a block in the current image; Between the horizontal profile Href (i) and the vertical profile Vref (i) for the block in the reference image and the horizontal profile Hcur (i) and the vertical profile Vcur (i) for the block in the current image And a motion vector estimator for calculating a difference value and estimating a coordinate difference of blocks having the calculated difference value as a motion vector.

Here, the block profile calculation unit may calculate a horizontal profile (Hcur (i)) and a vertical profile (Hcur (i)) for a block of the current image, (Vcur (i)) can be calculated using the following equation.

[Mathematical Expression]

Href (x) = Href (x + X, y + Y + i)

Vref (i) = Vref (x + X + i, y + Y) - Href (x + X + i, y +

Hcur (i) = Hcur (x, y + i) - Hcur (x-N, y + i)

Vcur (i) = Vcur (x + i, y) - Hcur (x + i,

(X, Y) is the motion vector, and N is the pixel size in the horizontal and vertical directions of the block)

The motion vector estimation unit may calculate a horizontal profile Hcur (i) and a vertical profile Vcur (i) for the block of the current image and a horizontal profile Hcur (i)) can be calculated as a Sum of Absolute Difference (SAD) or a Sum of Square Difference (SSD) value using the following equation.

[Mathematical Expression]

SAD = {? | Href (i) - Hcur (i) | +? | Vref (i) - Vcur (i) |} x?

SSD = {? | Href (i) - Hcur (i) | ² +? Vref (i) - Vcur (i) | ² } x?

(Where N is a scale constant and N is a natural number)

As described above, the profile information-based motion estimation apparatus and method according to the present invention generate a profile image for a reference image and a current image when estimating a motion vector for block matching, The motion vector is predicted by comparing only the profile information of the motion vector with respect to the entirety of the motion vector, thereby reducing the amount of calculation for motion prediction, thereby enabling high-speed image encoding.

FIG. 1 is a diagram illustrating a motion vector prediction state using a block matching method,
FIG. 2 is a control block diagram of a profile information-based motion estimation apparatus according to the present invention.
3 is a flowchart of a profile information-based motion estimation method according to the present invention.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

2 is a control block diagram of a profile information-based motion estimation apparatus according to the present invention.

2, the motion estimation apparatus of the present invention includes a reference image buffer 100, a current image buffer 110, a horizontal profile image generating unit 120, a vertical profile image generating unit 130, A calculation unit 140, a motion vector estimation unit 150,

The reference image buffer 100 outputs a reference image signal, and the current image buffer 110 outputs a current image signal. The reference image refers to an image temporally prior to the current image, or a previously processed image in the encoding / decoding order.

The horizontal profile image generating unit 120 generates horizontal profile images for the reference image and the current image, respectively. The horizontal profile is one-dimensional information calculated by summing the pixel values of the same horizontal line in the two-dimensional image of the image.

The horizontal profile image generation unit 120 sums the pixel values on the same horizontal line with respect to each horizontal line of the reference image output from the reference image buffer 100, and outputs the sum as a horizontal profile of the reference image.

Also, the horizontal profile image generation unit 120 sums the pixel values on the same horizontal line with respect to each horizontal line of the current image output from the current image buffer 110, and outputs the sum as a horizontal profile of the current image.

The vertical profile image generation unit 130 generates vertical profile images for the reference image and the current image, respectively. The vertical profile is one-dimensional information calculated by summing the pixel values of the same vertical line in the two-dimensional image of the image.

The vertical profile image generation unit 130 sums the pixel values on the same vertical line for each vertical line of the reference image output from the reference image buffer 100 and outputs the sum as a vertical profile of the reference image.

Also, the vertical profile image generator 130 sums the pixel values on the same vertical line for each vertical line of the current image output from the current image buffer 110, and outputs the sum as a vertical profile of the current image.

The following equations are used to generate the horizontal profile image and the vertical profile image.

[Equation 1]

H (X, y) = Σ X I (x, y)

&Quot; (2) "

V (x, Y) = Σ Y I (x, y)

Where (x, y) is the coordinates of the starting pixel of the block, (X, Y) is the motion vector, and N is the pixel size in the horizontal and vertical directions of the block.

The block profile calculation unit 140 calculates a horizontal profile and a vertical profile of a current image and a block in the reference image.

The block profile calculation unit 140 calculates the block profile of the reference image using the horizontal profile of the reference image output from the horizontal profile image generation unit 120 and the vertical profile of the reference image output from the vertical profile image generation unit 130, Lt; / RTI >

The block profile calculation unit 140 calculates a horizontal profile and a vertical profile of a block in the reference image for each candidate motion vector in the motion search area using Equations (3) and (4) below.

&Quot; (3) "

Href (x) = Href (x + X, y + Y + i)

&Quot; (4) "

Vref (i) = Vref (x + X + i, y + Y) - Href (x + X + i, y +

The block profile calculating unit 140 may calculate the current profile of the current image using the horizontal profile of the current image output from the horizontal profile image generating unit 120 and the vertical profile of the current image output from the vertical profile image generating unit 130, The horizontal and vertical profiles of the blocks are calculated.

The block profile calculation unit 140 calculates a horizontal profile and a vertical profile of the current intra-image block for each candidate motion vector in the motion search area using Equations (5) and (6) below.

&Quot; (5) "

Hcur (i) = Hcur (x, y + i) - Hcur (x-N, y + i)

&Quot; (6) "

Vcur (i) = Vcur (x + i, y) - Hcur (x + i,

The motion vector estimation unit 150 calculates motion prediction information, for example, a sum of absolute difference (SAD) or a sum of square difference (SSD), using the calculation result of the block profile. The motion vector estimator 130 estimates a candidate motion vector MV corresponding to a minimum cost among the costs corresponding to each candidate motion vector of the corresponding block as a final motion vector MV of the corresponding block. The motion vector estimator 150 may calculate SAD and SSD by applying the horizontal / vertical profile of the current block and the horizontal / vertical profile of the reference block to Equation (7) and Equation (8).

&Quot; (7) "

SAD = {? | Href (i) - Hcur (i) | +? | Vref (i) - Vcur (i) |} x?

&Quot; (8) "

SSD = {? | Href (i) - Hcur (i) | ² +? Vref (i) - Vcur (i) | ² } x?

Where alpha is a scale constant and can be set to N / 2 (N = natural number).

The motion vector estimating unit 150 estimates a coordinate difference between a region where the SAD or the SSD is minimized as a motion vector using Equations (7) and (8).

The motion vector storage unit 160 stores the estimated motion vectors.

As described above, in the present invention, if a profile image is generated once for the reference image and the current image, it can be applied to all the blocks in common, and only the profile information is compared with the candidate motion vector of each block (2NxN-1) operations are reduced to 4N at the time of NxN block comparison, resulting in a reduction in computation time.

For example, if N is 8, 127 operations are required according to the prior art, but in the case of the present invention, the number of operations is reduced to 32, so that the proposed method achieves a speed improvement of about 4 times.

3 is a flowchart of a profile information-based motion estimation method according to the present invention.

First, the reference image and the current image are input through the reference image buffer 100 and the current image buffer 110 (S110). The reference image refers to an image temporally previous to the current image or an image processed before encoding / decoding order, and each image is divided into macroblocks.

A horizontal profile image and a vertical profile image of the reference image are generated (S112). The horizontal / vertical profile image generating units 120 and 130 may apply Equation (1) to generate the horizontal profile image of the reference image and Equation (2) to generate the vertical profile image.

A horizontal profile image and a vertical profile image of the current image are generated (S114). The horizontal / vertical profile image generators 120 and 130 may apply Equations (1) and (2) to generate the horizontal profile image and the vertical profile image of the current image.

The block profile is calculated for the blocks in the reference image. Here, the reference image block profile includes a horizontal profile and a vertical profile of the reference image block (S116). Accordingly, the block profile calculation unit 140 can calculate the reference image block profile by applying Equations (3) and (4).

The block profile is calculated for the current block in the image. Here, the current image block profile includes a horizontal profile and a vertical profile of the current image block (S118). Accordingly, the block profile calculation unit 140 may calculate the current image block profile by applying Equations (5) and (6).

An error value between the reference image block profile and the current image block profile is calculated (S120). The motion vector estimator 150 calculates a matching error value and estimates a coordinate difference between the reference block and the current block having the minimum error value as a motion vector. To calculate the error value, SAD can be calculated using Equation (7) or SSD can be calculated using Equation (8).

Thus, those skilled in the art will appreciate that the present invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof. It is therefore to be understood that the embodiments described above are to be considered in all respects only as illustrative and not restrictive. The scope of the present invention is defined by the appended claims rather than the detailed description and all changes or modifications derived from the meaning and scope of the claims and their equivalents are to be construed as being included within the scope of the present invention do.

The present invention generates a profile image for a reference image and a current image when estimating a motion vector for block matching and compares only the profile information of the candidate motion vector of each block, The present invention can be applied to a profile information-based motion estimation apparatus and method for reducing a calculation amount for motion prediction and enabling high-speed encoding.

100: Reference image buffer
110: current image buffer
120: horizontal profile image generating unit
130: Vertical profile image generating unit
140: block profile calculation unit
150: motion vector estimator
160: a motion vector storage unit

Claims (6)

(a) a horizontal profile image generating unit and a vertical profile image generating unit receiving a reference image, which is a previous view image, from a reference image buffer and receiving a current image, which is a current view image, from a current image buffer;
(b) generating a horizontal profile of the reference image and a horizontal profile of the current image by summing pixel values on the same horizontal line for each horizontal line of the reference image and the current image, respectively;
(c) generating a vertical profile of a reference image and a vertical profile of a current image by summing pixel values on the same vertical line for each vertical line of the reference image and the current image, respectively;
(d) calculating a horizontal profile (Href (i)) and a vertical profile (Vref (i)) of the block in the reference image using the horizontal profile and the vertical profile of the reference image, Calculating a horizontal profile (Hcur (i)) and a vertical profile (Vcur (i)) for the block in the current image using the horizontal profile and the vertical profile;
(i) a vertical profile (Href (i)) and a horizontal profile (Hcur (i)) for a block in the current image, and a vertical profile (I)) and the horizontal profile (Hcur (i)) for the block of the current image, using the horizontal profile (Href (i)) and the vertical profile (Vref And a vertical profile (Vcur (i)), respectively;
(i)) and a horizontal profile (Hcur (i)) of the block of the current image calculated by the motion vector estimation unit and a horizontal profile (Href (i)) and a vertical profile Estimating a coordinate difference of blocks having a minimum difference value between the profiles (Vcur (i)) as a motion vector;
Lt; / RTI >
The reference image and the current image are divided into macroblocks,
The step (b)
The horizontal profile of the reference image and the horizontal profile of the current image
"H (X, y) = ^X I (x, y)"
, ≪ / RTI >
The step (c)
The vertical profile of the constant reference image and the vertical profile of the current image
"V (x, Y) = Σ Y I (x, y)"
, ≪ / RTI >
The step (d)
A horizontal profile Hcur (i) and a vertical profile Vcur (i) for the block of the current image and a horizontal profile Hcur (i) and a vertical profile Vcur Wherein the motion vector is calculated using the following equation.
[Mathematical Expression]
Href (x) = Href (x + X, y + Y + i)
Vref (i) = Vref (x + X + i, y + Y) - Href (x +
Hcur (i) = Hcur (x, y + i) - Hcur (xN, y + i)
Vcur (i) = Vcur (x + i, y) - Hcur (x + i, yN)
(X, Y) is the motion vector, and N is the pixel size in the horizontal and vertical directions of the block)

delete ◈ Claim 3 is abandoned due to the registration fee. The method according to claim 1,
The step (e)
Between the horizontal profile Href (i) and the vertical profile Vref (i) of the block of the reference image and the horizontal profile Hcur (i) and the vertical profile Vcur (i) Wherein the difference value is calculated as a Sum of Absolute Difference (SAD) or a Sum of Square Difference (SSD) value using the following equation.
[Mathematical Expression]
SAD = {? | Href (i) - Hcur (i) | +? | Vref (i) - Vcur (i) |} x?
SSD = {? | Href (i) - Hcur (i) | ² +? Vref (i) - Vcur (i) | ² } x?
(Where N is a scale constant and N is a natural number)
A reference image buffer for outputting a reference image signal which is a video image at a previous time point;
A current image buffer for outputting a current image signal which is a current time image;
A horizontal profile that receives the reference image and the current image and generates a horizontal profile of the reference image and a horizontal profile of the current image by summing the pixel values on the same horizontal line for each horizontal line of the reference image and the current image, An image generation unit;
A vertical profile that receives the reference image and the current image and generates a vertical profile of the reference image and a vertical profile of the current image by summing pixel values on the same vertical line for each vertical line of the reference image and the current image, An image generation unit;
(Href (i)) and a vertical profile (Vref (i)) for a block in the reference image using the horizontal profile and the vertical profile of the reference image, and using the horizontal profile and the vertical profile of the current image A block profile calculation unit for calculating a horizontal profile (Hcur (i)) and a vertical profile (Vcur (i)) for a block in the current image;
Between the horizontal profile Href (i) and the vertical profile Vref (i) for the block in the reference image and the horizontal profile Hcur (i) and the vertical profile Vcur (i) for the block in the current image A motion vector estimator for calculating a difference value and estimating a coordinate difference of blocks having the calculated difference value as a motion vector;
Lt; / RTI >
The reference image and the current image are divided into macroblocks,
The horizontal profile image generation unit
The horizontal profile of the reference image and the horizontal profile of the current image
"H (X, y) = ^X I (x, y)"
, ≪ / RTI >
The vertical profile image generation unit
The vertical profile of the constant reference image and the vertical profile of the current image
"V (x, Y) = Σ Y I (x, y)"
, ≪ / RTI >
The block profile calculation unit
A horizontal profile Hcur (i) and a vertical profile Vcur (i) for the block of the current image and a horizontal profile Hcur (i) and a vertical profile Vcur Wherein the motion vector is calculated using the following equation.
[Mathematical Expression]
Href (x) = Href (x + X, y + Y + i)
Vref (i) = Vref (x + X + i, y + Y) - Href (x +
Hcur (i) = Hcur (x, y + i) - Hcur (xN, y + i)
Vcur (i) = Vcur (x + i, y) - Hcur (x + i, yN)
(X, Y) is the motion vector, and N is the pixel size in the horizontal and vertical directions of the block)
delete ◈ Claim 6 is abandoned due to the registration fee. 5. The method of claim 4,
The motion vector estimation unit
Between the horizontal profile Href (i) and the vertical profile Vref (i) of the block of the reference image and the horizontal profile Hcur (i) and the vertical profile Vcur (i) Wherein the difference value is calculated as a Sum of Absolute Difference (SAD) or a Sum of Square Difference (SSD) value using the following equation.
[Mathematical Expression]
SAD = {? | Href (i) - Hcur (i) | +? | Vref (i) - Vcur (i) |} x?
SSD = {? | Href (i) - Hcur (i) | ² +? Vref (i) - Vcur (i) | ² } x?
(Where N is a scale constant and N is a natural number)

Images