KR100814715B1 - Moving-picture coding apparatus and method - Google Patents

Abstract

Disclosed are a video encoder, a decoder, and a method thereof. The video encoding apparatus of the present invention, when there is camera movement, compensates for global movement of the entire screen by the movement of the camera from a reference video signal using camera motion parameters, and outputs a motion compensated image. Compensation unit; A subtractor for outputting a difference between a current video signal and the motion compensated video; A conversion / quantization unit for converting and quantizing the difference into a frequency band; And an encoding unit for encoding the quantized result. According to the present invention, a much higher compression ratio can be obtained at the same image quality as compared with the conventional method.

Description

Video encoder, decoder and method {Moving-picture coding apparatus and method}

1 is a block diagram of a video encoder according to an embodiment of the present invention.

2 is a block diagram of a video decoder according to an embodiment of the present invention.

3 is a flowchart illustrating a global motion estimation and compensation method according to an embodiment of the present invention.

4 is a diagram illustrating a change in pixel coordinate when there is camera movement.

5 is a flowchart illustrating detailed processes of the global motion parameter estimation process using the RTLS algorithm according to an embodiment of the present invention.

The present invention relates to a video compression encoding method and apparatus, and more particularly, to a method and apparatus for compensating / decoding a motion of an entire screen by a camera.

Recently, products and communication media for digital video services have been rapidly developed. Among them, video application technologies using video standardization standards such as MPEG or H.264 are providing video services to homes through ATM (Asynchronous Transfer Mode) or over-the-air communication. In order to provide various video services through current communication media, an efficient video compression encoding method and apparatus are needed. In particular, the H.264 compression coding scheme has been proposed in various compression schemes to provide better image quality by increasing the compression rate compared to the MPEG-2 scheme. See, eg, "Overview of the H.264 / AVC Video Coding Standard" by T. Wiegand et al., IEEE trans. on Circuits and Systems for Video Tech., vol. 13, no. 7, pp. 560-576, July 2003, and the like.

The H.264 / AVC encoder according to the prior art is composed of an intra-frame predictive encoder, an inter-frame predictive encoder, a transform encoding, a quantization and an entropy encoder, and the like. Through this, encoding is performed with high compression. In particular, in the case of the motion compensation circuit, only a temporal change between the previous frame and the current frame is estimated and compensated for, thereby reducing a large amount of information. Therefore, inter-frame motion estimation is a very important technique of high-performance video compression coding.

MPEG compression coding schemes including H.264 divide a frame image into block units for motion estimation and perform a motion estimation technique for each block. The amount of coordinate change between pixels in up, down, left, and right in relation to the motion of each block unit is called a motion vector, and one motion vector determines the motion change of the block.

If the camera moves, the change between frames due to camera movement is seen globally rather than locally. Therefore, if you find a parameter representing the overall motion of the screen and send only this parameter, the screen can be reconstructed in a very small amount.

In the case of pan-tilt-zoom (PTZ) CCTV cameras used in digital video recorders (DVR) or IP camera systems, the camera zooms in and out and rotates in three dimensions. In such an application system, if the camera parameter corresponding to the three-dimensional motion of the camera is used to compensate for this, a very small amount of data can compensate for a considerable amount of motion.

An object of the present invention is to provide a video encoding / decoding apparatus and method capable of obtaining high-efficiency compression performance by compensating for global motion of an entire screen using motion parameters of a camera.

In order to achieve the above technical problem, the video encoding apparatus according to an aspect of the present invention estimates global motion of the entire screen using a previously stored reference video signal and a current video signal only when there is a camera movement based on a flag. A global motion estimator for calculating global motion parameters; A global motion compensator configured to compensate for the global motion by using the global motion parameters and to output an image compensated for the global motion; A local motion estimator for estimating local motion of the current video signal; A local motion compensator for outputting a motion compensated image by compensating the local motion with respect to the image compensated for the global motion based on an estimation result of the local motion estimator; Receives a flag indicating whether the camera is output from the camera control unit for controlling the camera, and selectively transmits the current video signal output from the camera in response to the flag to the global motion estimation unit or the local motion estimation unit Providing switch; A subtractor for outputting a difference between the current video signal and the motion compensated video; A conversion / quantization unit for converting and quantizing the difference into a frequency band; And an encoding unit for encoding the quantized result. The global motion estimator receives the current video signal, selects a plurality of feature points, compares the current video signal with the reference video signal, and matches the plurality of feature points with respect to each of the plurality of feature points. Extract second pixel coordinate values in the current image signal corresponding to first pixel coordinate values in a reference image signal, and estimate the global motion parameters using the first and second pixel coordinate values. Can be.

In order to achieve the above technical problem, a video decoding apparatus according to an aspect of the present invention uses a global motion parameter received from an image encoding device to compensate for a global motion of an entire screen by a camera motion from a reference video signal, A motion predictor / compensator for outputting a compensated image; A decoder which decodes an encoded video signal; An inverse quantization / inverse transform unit for inversely quantizing an output signal of the decoder and inversely converting the inverse quantized result into a time band; And an adder for adding the output signal of the inverse quantization / inverse transform unit with the motion compensated image signal.

According to an aspect of the present invention, there is provided a video encoding / decoding method, which receives a flag indicating whether the camera is output from a camera controller for controlling a camera, and receives a flag from the camera in response to the flag. Selectively providing an output current video signal to a global motion estimator or a local motion estimator; A global motion estimation step of calculating global motion parameters by estimating global motion of the entire screen using a pre-stored reference video signal and the current video signal only when there is motion of the camera based on the flag; A global motion compensation step of compensating for the global motion by using the global motion parameters and outputting an image in which the global motion is compensated; A local motion estimation step of estimating local motion of the current video signal; A local motion compensation step of outputting a motion compensated image by compensating the local motion with respect to the image compensated for the global motion based on an estimation result of the local motion estimator; Outputting; Converting and quantizing the difference into a frequency band; And encoding and outputting the quantized result.

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

1 is a block diagram of a video encoder according to an embodiment of the present invention.

The encoder 100 as shown is a transform / quantizer 110, inverse quantization / inverse transform unit 120, the first and second adders (130, 135), de-blcoking unit 140, The storage unit 150 includes an intra frame predictor 160, a motion compensator 170, a motion estimator 180, an encoder 190, and a camera controller 195.

The illustrated video encoder 100 processes an input image in units of macro blocks having a predetermined pixel size (for example, 16 × 16 pixels), and several macro blocks are bundled in slice units. According to the image size, one frame may include several slices.

First, an operation in which intra coding is performed will be described.

The first adder 130 outputs a difference between the input macro block and the prediction macro block output from the intra frame predictor 160. To this end, the switch SW1 is switched so that the output of the intra frame predictor 160 is provided to the first adder 130. The intra frame predictor 160 generates a prediction macro block by using a macroblock encoded and reconstructed with respect to a previous frame stored in the storage 150, and outputs the generated prediction macroblock. A detailed description of the operation of the intra frame predictor 160 will be described later.

The transform / quantization unit 110 converts the difference between the current macroblock and the prediction macroblock output from the first adder 130 into a frequency band signal through the DCT, and quantizes the transformed DCT coefficients. The quantized DCT coefficients output from the transform / quantization unit 110 are input to the inverse quantization / inverse transform unit 120 and the encoder 190.

The encoder 190 outputs the quantized DCT coefficients by entropy coding according to, for example, variable length coding or context-adaptive VLC.

The inverse quantization / inverse transform unit 120 inverse quantizes and inverses DCT the quantized DCT coefficients. The second adder 135 adds the inverse DCT result, that is, the output data of the inverse quantization / inverse transformer 120 with the prediction macro block output from the intra frame predictor 160. In this case, since the image is blocked in units of macroblocks and includes blocking due to quantization differences between macroblocks, deblocking filtering is performed by the deblocking unit 140 to remove the blocking effect. The final filtered and reconstructed macroblock is stored in the frame storage unit 150 and used for prediction of the next macroblock.

Next, an operation in which inter coding is performed is as follows.

The motion estimation unit 180 estimates the most similar macroblock among the current macroblock and reference frames previously stored in the frame storage unit 150. In this case, the reference frame refers to pictures previously encoded and stored for motion estimation. Here, the reference frame may be an I-picture or a P-picture in the case of MPEG-2, and a B-picture as well as an I-picture and a P-picture in H.264. (Bi-directional predictive-picture)

 The motion estimator 180 includes a global motion estimator 181 and a local motion estimator 182. The global motion estimator 181 estimates global motion of the entire screen due to the movement of the 3D camera 105 from the reference frame stored in the storage 150 and the current frame, and thus the global motion parameters (camera motion parameters). Outputs The global motion parameters are camera motion parameters (hereinafter referred to as first) consisting of a zoom factor (s), a pan angle, a tilt angle, a swing angle and a focal length. Camera motion parameter set). Or, the global motion parameters are based on five camera motion parameters based on a zoom factor s, a pan angle, a tilt angle, a swing angle and a focal length. The parameters may be obtained (hereinafter also referred to as a second camera motion parameter set). These global motion parameters are transmitted to the video decoder (200 of FIG. 2) through the encoder 190.

The local motion estimator 182 estimates a local motion (eg, a motion caused by the movement of a person or an object) generated locally on a portion of the screen. Specifically, the local motion estimation unit 182 performs motion estimation of an input macro block with respect to a reference frame and outputs motion data including a motion vector and an index indicating a reference frame.

The motion compensator 170 generates and outputs a macro block whose motion is predicted from the reference frame used for the motion estimation using the estimation result of the motion estimator 180. The macro block thus generated is called a motion compensated macro block. In detail, the motion compensator 170 may include a global motion compensator (not shown) and a local motion compensator (not shown). The global motion compensator (not shown) generates an image (macro block) compensated for the global motion from the reference frame based on the global motion estimation result of the global motion estimator 181, that is, the global motion parameters. The local motion compensator (not shown) compensates the local motion from the image in which the global motion is compensated based on the local motion estimation result in the local motion estimator 182, so that the final compensation image, that is, both the global motion and the local motion, is The compensated image (macro block) is output.

The global motion estimator 181 and the compensator may operate only when there is global motion by the 3D camera 105. In the case of a video system using a 3D camera such as a security video system, the 3D camera 105 may be zoom-in / zoom-out, pan, or tilt under the control of the camera controller 195. Perform three-dimensional movements, such as tilt. In this case, the camera controller 195 may determine whether the camera 105 moves, and output a flag CSW2 indicating whether the camera moves. Therefore, when the flag CSW2 output from the camera controller 195 is activated, that is, when there is camera movement, the global motion estimator 181 and the compensator may be operated.

In this embodiment, the switch SW2 selectively provides the input signal to the global motion estimation unit 181 and the local motion estimation unit 182 in response to the flag CSW2. Accordingly, when the flag CSW2 is activated, that is, when there is camera movement, the input signal is input to the global motion estimation unit 181, so that global motion estimation is performed first and local motion is estimated. On the other hand, when the flag CSW2 is not activated, that is, when there is no camera motion, the input signal is input to the local motion estimator 182 to perform only local motion estimation.

The switch SW1 is switched so that the output of the motion compensator 170, that is, the motion compensated macro block is provided to the first adder 130.

The first adder 130 subtracts the motion-compensated macroblock from the input macroblock to obtain the difference, and the transform / quantization unit 110 performs DCT and quantization on the output of the first adder 130, 190 entropy-codes the quantized DCT coefficient and the motion vector estimated by the motion estimation unit 180 to output a compressed bit stream.

The first camera motion parameter set (zoom factor, pan angle, tilt angle, swing angle, and focal length) may be output from the camera controller 195. That is, the camera controller 195 may provide the first camera motion parameter set to the global motion estimator 181 or the global motion compensator (not shown). In this case, the process of estimating the global motion parameter by the global motion estimation unit 181 may be omitted or simplified.

2 is a block diagram of a video decoder 200 according to an embodiment of the present invention. The video decoder 200 according to an exemplary embodiment of the present invention includes a buffer 210, a decoder (decoder 215), an inverse quantizer 220, a buffer 225, an inverse transform unit 230, and an adder 235. , A de-blcoking unit 240, a storage unit 245, an intra frame predictor 250, a global motion compensator 255, and a local motion compensator 260. This configuration has a structure in which the global motion compensation unit 255 is added as compared to the conventional H.264 video decoder.

First, an operation in which intra decoding is performed is as follows.

The buffers 210 and 225 temporarily store input signals.

An input signal received from the video compression encoder 100, that is, an entropy coded signal (bit stream), is input to the decoder 215 through the buffer 210. The decoder 215 decodes the entropy coded bitstream and outputs quantized DCT coefficients.

The inverse quantizer 220 inversely quantizes an output signal of the decoder 215, that is, a quantized DCT coefficient, and outputs the inverse quantized coefficient to the buffer 225. The output signal of the buffer 225 is input to the inverse transformer 230, and the inverse transformer 230 performs inverse DCT. The inverse quantizer 220 and the inverse transformer 230 are substantially the same as the functions of the inverse quantizer / inverse transformer 120 of the video encoder 100.

The adder 235 adds the inverse DCT result data to the prediction macro block output from the intra frame predictor 250. To this end, the switch SW3 is switched so that the output of the intra frame predictor 250 is input to the adder 235.

The output data of the adder 235 is deblocked filtered by the deblocking unit 240. The final filtered and reconstructed macroblock is stored in the storage unit 245 and is also used for prediction of the next macroblock. The deblocking unit 240 and the intra frame predicting unit 250 are substantially the same as the functions of the deblocking unit 140 and the intra frame predicting unit 160 of the video encoder 100, respectively.

 Next, an operation in which inter decoding is performed is as follows.

In the inter decoding operation, an input signal passes through the buffer 210, the decoder 215, the inverse quantization unit 220, the buffer 225, and the inverse transformer 230. Since the same, detailed description thereof will be omitted.

The adder 235 adds the inverse DCT result data to the macro block compensated for the motion output from the local motion compensator 260. To this end, the switch SW3 is switched so that the output of the local motion compensator 260 is input to the adder 235.

The global motion compensator 255 reconstructs the image in which the global motion is compensated from the reference frame by using the camera motion parameters transmitted from the video encoder 100. More specifically, when there is a camera motion, the global motion compensator 255 uses the camera motion parameters to obtain pixel values corresponding to the pixel coordinates X and Y in the previous frame, as shown in Equation 1 below. Bring to the predicted pixel coordinates (X ', Y') in the current frame. At this time, since (X ', Y') may be real coordinates, pixel values of integer coordinates are predicted using interpolation (for example, bilinear interpolation).

If there is a global motion, the local motion compensator 260 receives an image compensated for the global motion from the global motion compensator 255, and reconstructs the image for which the local motion is compensated. To this end, the switch SW4 is switched so that the output of the storage unit 245 is input to the global motion compensation unit 255. Therefore, the output of the local motion compensator 260 is an image reconstructed by compensating for the global motion and the local motion.

When there is no global motion, the output of the storage unit 245 is input to the local motion compensator 260 without passing through the global motion compensator 255. To this end, the switch SW4 is switched so that the output of the storage unit 245 is input to the local motion compensation unit 260.

The output data of the adder 235 is stored in the storage unit 245 after deblocking filtering is performed by the deblocking unit 240.

As described above, the video encoder / decoder according to an embodiment of the present invention further compensates local movements on the screen that compensates for the change of the overall screen changed by the camera movement, thereby obtaining a good image quality. It becomes possible.

3 is a flowchart illustrating a global motion estimation and compensation method according to an embodiment of the present invention. The global motion estimation and compensation method according to an embodiment of the present invention may be performed by the global motion estimation unit 181 and the motion compensation unit 170 of the video encoder.

Referring to Figures 1 and 3, the global motion estimation and compensation method according to an embodiment of the present invention will be described.

First, a plurality of feature points (eg, N and N are integers of 2 or more) are selected by receiving the first image signal (S10). The first image signal is a current input image and may be a current frame or a current macro block. The feature point is a partial image that is a feature of the first image signal, and is a part of an image including an edge or a high texture. For example, in an image signal in which a person appears, a human eye, a nose, a mouth, and the like may be characteristic points.

Next, a matching process for the selected feature points is performed by comparing the first image signal and the second image signal (S20). For feature point matching, a feature point matching algorithm such as a block matching algorithm may be used. The second video signal may be a previous video signal or a reference video signal of the first video signal.

The pixel coordinate values X 'and Y' of the first image signal corresponding to the pixel coordinate values X and Y of the feature points of the second image signal may be obtained through the feature point matching process.

When there is camera movement, the pixel coordinates X 'and Y' in the current frame corresponding to the pixel coordinates X and Y in the previous frame may be obtained using camera movement parameters.

4 is a diagram illustrating a change in pixel coordinate when there is camera movement. Referring to FIG. 4, a pan angle α, a tilt angle β, a swing angle γ are given, and a focal length F is F ₁ to F _2. When changed to, the pixel coordinates X and Y are changed to the new pixel coordinates X 'and Y'. In this way, the relationship between the pixel coordinates (X, Y) in the previous frame and the pixel coordinates (X ', Y') in the current frame when there is camera movement is expressed by Equation 1 below.

Where F ₁ and F ₂ are the focal lengths of the camera before and after zoom, respectively. The zoom factor s is F ₂ / F ₁ . And r _ij (i, j = 1,2,3) are the elements of the 3D rotation matrix R. Parameters of Equation 1 include five 3D camera motion parameters: zoom factor (s), pan angle (α), tilt angle (β), swing angle (γ) and focal length (focal length, F).

Next, sets of pixel coordinates X and Y in the previous frame (second image signal) and pixel coordinates X 'and Y' in the current frame (first image signal) for the feature points obtained through feature point matching ( sets), a global motion parameter estimation process is performed (S30).

Global motion parameter estimation may be performed by the global motion estimation unit 181 using a recursive total least mean squares (RTLS) algorithm.

Detailed processes for the global motion parameter estimation process S30 using the RTLS algorithm are described in FIG. 5.

)로부터 q벡터(

)를 구한다(S110). 여기서, a 벡터(

)는 여덟 개의 움직임 파라미터들(제2 카메라 움직임 파라미터 셋)로서, 상술한 5개의 카메라 움직임 파라미터들(제1 카메라 움직임 파라미터 셋)로부터 다음의 수학식 2에 의해 구해질 수 있다. Referring to this, first, a vector (a permutation)

From q vector (

To obtain (S110). Where a vector (

) Is eight motion parameters (second camera motion parameter set), and can be obtained from Equation 2 below from the five camera motion parameters (first camera motion parameter set) described above.

)는 다음의 수학식 3과 같이 정의될 수 있다.q vector (

) May be defined as in Equation 3 below.

)와

를 임의의 값(

,

)으로 초기화한다(S120).Next, q vector (

)Wow

Is a random value (

,

Initialize to) (S120).

The variables k and m are initialized to 1, respectively (S130).

)를 선택한다(S140). 입력 벡터(

)는 다음의 수학식 4에 의해 구해질 수 있다.Then, input vector (

Select (S140). Input vector (

) Can be obtained by the following equation (4).

here,

를 업데이트한다(S150).

업데이트에 적용되는 수식은 다음의 수학식 5와 같다. Next, we use the matrix inversion lemma

Update (S150).

The equation applied to the update is shown in Equation 5 below.

here,

는 이전 값으로부터 간단하게 업데이트될 수 있다.therefore,

Can simply be updated from the previous value.

Next, α is calculated using Equation 6 (S160).

here,

ego,

D is a 9 x 9 symmetric non-negative matrix.

를 이용하여 다음의 수학식 7과 같이 q벡터(

)를 업데이트한다(S170).Next, with the calculated α,

By using the q vector (

) Is updated (S170).

Steps S140 to S170 are repeatedly performed for m = 1,2 from k = 1 to N (number of selected feature points). That is, after step S170, check whether m = 2 (S180), if m = 2, add m by one (S200), return to step S140, and check whether k = N (S190) and k = If not N, k is added (S210), and the process returns to step S140.

)와로부터 a벡터(

)를 산출한다. Finally, the q vector (

) And a vector (

) Is calculated.

Referring back to FIG. 3, after the global motion parameter estimation process S30 is performed, it is determined whether there is a local motion (S40), and if there is a local motion, the local motion is removed and the global motion parameter estimation process S30 is performed again. Perform. If there is no local motion, the next step three-dimensional analysis and global motion compensation step (S60) is performed. Finding local motion and removing local motion is to prevent the global motion parameter due to local motion from being incorrectly estimated. That is, in the global motion parameter estimating step, the local motion is estimated as much as possible to estimate the global motion parameter.

)를 이용하여 글로벌 움직임이 보상된 영상을 생성한다.In the global motion compensation step S60, the estimation result in the global motion parameter estimating process S30, that is, the a vector (the estimated global motion parameters) (

) To generate an image compensated for global motion.

Since the video compression encoding method using the camera motion parameter of the present invention as a whole compensates for the three-dimensional motion of the camera, it is possible to obtain a much higher compression rate at the same image quality as compared with the conventional method. For example, when the camera zooms in, the conventional methods estimate the motion on a block-by-block basis and send information of different motion vectors, whereas the method according to the present invention is very efficient because the screen can be configured by transmitting only zoom parameters. .

The invention can also be embodied as computer readable code on a computer readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data that can be read by a computer system is stored. Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage, and the like, and also implemented in the form of a carrier wave (for example, transmission over the Internet). It includes being. The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion. And functional programs, codes and code segments for implementing the present invention can be easily inferred by programmers in the art to which the present invention belongs.

Although the present invention has been described with reference to the above embodiments, it is merely illustrative, and those skilled in the art will understand that various modifications and equivalent other embodiments are possible therefrom. . Therefore, the true technical protection scope of the present invention will be defined by the technical spirit of the appended claims.

According to the present invention, since the global motion of the entire screen due to the three-dimensional motion of the camera is compensated, it is possible to obtain a much higher compression ratio in the same image quality compared to the conventional method. In addition, when performing the global motion compensation according to the present invention, when compared to the prior art, even when the three-dimensional motion of the camera, there is much better predictive coding performance can be obtained. Therefore, when the present invention is applied to a security video system or a video system such as DTV, CCTV, the performance and quality of these video systems can be improved.

Claims (13)

A global motion estimator for estimating global motion of the entire screen by using a previously stored reference video signal and a current video signal based on a flag, and calculating global motion parameters; A global motion compensator configured to compensate for the global motion by using the global motion parameters and to output an image compensated for the global motion; A local motion estimator for estimating local motion of the current video signal; A local motion compensator for outputting a motion compensated image by compensating the local motion with respect to the image compensated for the global motion based on an estimation result of the local motion estimator; Receives a flag indicating whether the camera is output from the camera control unit for controlling the camera, and selectively transmits the current video signal output from the camera in response to the flag to the global motion estimation unit or the local motion estimation unit Providing switch; A subtractor for outputting a difference between the current video signal and the motion compensated video; A conversion / quantization unit for converting and quantizing the difference into a frequency band; And And an encoding unit for encoding the quantized result. The method of claim 1, wherein the global motion estimation unit Receiving the current video signal to select a plurality of feature points, compare the current video signal and the reference video signal to match the plurality of feature points in the reference video signal for each of the plurality of feature points And extracting the second pixel coordinate values in the current image signal corresponding to the first pixel coordinate values of and estimating the global motion parameters using the first and second pixel coordinate values. The method of claim 2, The plurality of feature points include first to Nth (integer two or more) feature points, The global motion estimator Initialize a predetermined first vector (q vector) and use the first pixel coordinate values and the second pixel coordinate values with respect to a corresponding feature point among the first vector and the first to Nth feature points. Iteratively repeating the step of updating the first vector (qvector) with respect to the first to Nth feature point to calculate the global motion parameter from the last updated first vector (qvector) Device. The method of claim 3, wherein the global motion estimation unit  And calculating the global motion parameters after estimating the local motion, removing the estimated local motion. The method of claim 3, wherein a relationship between the second vector (a vector) and the first vector (q vector) including the global motion parameters is

(here,

Is the vector,

Is a vector, a ₁ to a ₈ are the global motion parameters, A ₁ to a ₈ are

Where F ₁ and F ₂ are the focal lengths before and after the camera movement, s is the zoom factor, and r _ij (i, j = 1,2,3) is the element of the three-dimensional rotation matrix as the zoom factor of the camera. (s), determined by camera movement parameters including pan angle (α), tilt angle (β), swing angle (γ) and focal length (F) Video encoding apparatus. The image encoding apparatus of claim 5, wherein the camera motion parameters are received from the camera controller. Receiving a flag indicating whether the camera is output from a camera controller for controlling a camera; Selectively providing a current video signal output from the camera to a global motion estimator or a local motion estimator in response to the flag; Only when there is motion of the camera based on the flag, the global motion estimator calculates global motion parameters by estimating global motion of the entire screen using a pre-stored reference video signal and the current video signal. Motion estimation step; A global motion compensation step of compensating for the global motion by using the global motion parameters and outputting an image in which the global motion is compensated; A local motion estimation step of estimating a local motion of the current video signal by the local motion estimator; A local motion compensation step of outputting a motion compensated image by compensating the local motion with respect to the image with the global motion compensated based on an estimation result of the local motion estimator; Outputting a difference between the current video signal and the motion compensated video; Converting and quantizing the difference into a frequency band; And And encoding and outputting the quantized result. The method of claim 7, wherein the global motion estimation step And estimating local motion of the entire screen and calculating the global motion parameters after removing the estimated local motion. The method of claim 8, wherein the global motion estimation step Selecting a plurality of feature points including first to Nth feature points from the current video signal; A second pixel coordinate in the current video signal corresponding to first pixel coordinate values in the reference video signal for each of the selected first to Nth feature points by comparing the current video signal with the reference video signal Extracting values; And Estimating the global motion parameters using the first pixel coordinate values and the second pixel coordinate values for each of the first to Nth feature points. 10. The method of claim 9, wherein estimating the global motion parameters Initializing a first vector (q vector); Updating the first vector (q vector) by using the first pixel coordinate values and the second pixel coordinate values of the first vector and the corresponding feature point among the first to Nth feature points. Iteratively repeating for the first to Nth feature points; And And calculating the global motion parameter from the last updated first vector q through the sequentially repeating steps. 11. The method of claim 10, wherein the relationship between the second vector (a vector) and the first vector (q vector), which are composed of the global motion parameters, is

(here,

Is the vector,

Is a vector, a ₁ to a ₈ are components of the a vector as the global motion parameters, A ₁ to a ₈ are

Where F ₁ and F ₂ are the focal lengths before and after the camera movement, s is the zoom factor, and r _ij (i, j = 1,2,3) are the elements of the three-dimensional rotation matrix, respectively. s), determined by camera movement parameters including pan angle (α), tilt angle (β), swing angle (γ) and focal length (F)) Video encoding / decoding method. The method of claim 7, wherein the video encoding / decoding method Transmitting the global motion parameters from an image encoding apparatus to an image decoding apparatus; And And decoding the received video signal by using the global motion parameters in the video decoding apparatus. A computer-readable recording medium having recorded thereon a program for realizing the video encoding / decoding method according to any one of claims 7 to 12.

Images