JP2962766B2 - Method for determining image scaling and translation parameters - Google Patents

Description

Description: BACKGROUND OF THE INVENTION The present invention relates to a method for determining a parameter that expresses enlargement / reduction and translation of an entire image for efficient encoding of an image signal.

(Prior Art) As a conventional technique for estimating a parameter expressing the movement of an entire image from local motion vectors in block units obtained for two digital image frames, for example, Japanese Patent Application No. Hei. There is 329734.

In Japanese Patent Application No. 1-329734, three types of parameters representing the movement of a camera due to panning and zooming are estimated based on a mean square error criterion using motion vectors obtained for each block. It is characterized in that it can be executed only by a sum operation. The parameter estimation method of Japanese Patent Application No. 1-329734 will be described below.

Parameters Is defined by the following equation.

Here, Z is a parameter representing the enlargement / reduction of the entire image caused by the zoom of the camera, H is a parameter representing the horizontal translation of the entire screen caused by the pan of the camera, and V is the parallelism of the entire screen in the vertical direction. A parameter representing the movement.

FIG. 4 shows an example of how to give a block and coordinates in an image. As shown in the figure, when the center (00) of the image is the origin and the size of each block is b ₁ × b ₂ , the block B _{i , j} are (ib ₁ , jb ₂ ). In the coordinate system of FIG. Is calculated by the following equation.

here, It is. Also, Is a motion vector in the block having the center coordinates (ib ₁ , jb ₂ ), and is represented by the following equation.

(W _h : the horizontal component of the motion vector) (w _v : the vertical component of the motion vector) Is a 3-by-3 matrix consisting of constant elements. Japanese Patent Application No. 1
In -329734, the parameter according to equation (2) Is calculated using all the motion vectors of the image blocks at once.

(Problems to be Solved by the Invention) As described above, the parameters are set based on the mean square error criterion using all the motion vectors at once. Is determined, if the screen contains local motion other than enlargement / reduction and parallel movement of the entire image, such as when the object in the screen itself is moving, it is caused by the local motion. The resulting motion vector is also used for determining the parameter, so that an error occurs in the value of the parameter.

(Object of the Invention) The present invention solves the above-mentioned problem, and does not cause an error in parameter values even when an image includes a local motion other than enlargement / reduction and parallel movement of the entire image. It is intended for.

(Means for Solving the Problems) In order to achieve the above object, the present invention provides a process of estimating parameters using a part of motion vectors from among motion vectors obtained for each block. Is performed a plurality of times while changing the motion vector used in the calculation, the estimated values of a plurality of parameters are obtained for the entire image, and the most frequently used estimated value among them is set as the parameter of the image.

(Operation) In areas where there is local motion other than enlargement / reduction and translation of the whole image, those motion vectors have various directions and magnitudes. Therefore, parameters estimated using these motion vectors have various values, and the frequency of each value is low. In the present invention, since the most frequent value is used as a parameter, a motion vector representing a local motion other than enlargement / reduction and parallel movement does not affect the parameter determination, and no error occurs in the parameter value.

(Embodiment) FIG. 1 shows an example of how to give blocks and coordinates in an image, where B _{i, j} has a block size of b ₁ × b ₂ and a center coordinate of (ib ₁ , jb ₂ ). Block. In FIG. 1, two blocks B symmetrical to each other with respect to i = 0
_{i, j,} B _-i, from motion vector of _j, the parameters Is estimated. Similarly, from the motion vectors of two blocks B _{i, j} , B _{i, −j} symmetric with respect to j = 0, the parameter Is estimated. here, And Z1 _{i, j} , H1 _{i, j} , V1 _{i, j} are two blocks B _{i, j} , B
_{−i, j} , Z2 _{i, j} , H2 _{i, j} , V
2 _{i, j} is a parameter value obtained by using two blocks B _{i, j} , B _{i, −j} .

Is a frequency distribution for each element. FIG. 2 is an example of a frequency distribution of parameter estimation values in each element, where
, That is, Z = 0.025... FIG. 2 (1) P _h = −2.5... FIG. 2 (2) P _v = 1.0... FIG.

FIG. 3 is a block diagram of a parameter determining circuit according to an embodiment for implementing the method of the present invention.

In the figure, 1, 11 are address data (i, j), (−
i, j), 2,12 are matrices Matrix respectively in the calculation circuit 3, matrix Outputs 13. Reference numerals 4 and 14 denote motion vector designating circuits for outputting motion vectors 7 and 17, respectively. A motion vector memory 5 outputs motion vectors 6 for all blocks. 8, 18 is the product-sum operation circuit 3 and 13 and the motion vectors 7 and 17 are input, and the product-sum operation values 9 and 19 are output. Reference numeral 10 denotes an adder for the product-sum operation values 9, 19, which outputs an addition value 20. 21 is a matrix The address data 1, 11 and the matrix The entire pattern 23 of the memory 22 is input, and only the matrix 24 determined by the address data is output. Reference numeral 25 denotes a product-sum operation circuit, a parameter 26 representing an enlargement / reduction of the image by performing a product-sum operation using the added value 20, the matrix 24 as an input, a parameter 27 representing a horizontal translation of the image, and a vertical direction of the image. The parameter 28 representing the parallel movement to is output. Reference numerals 29 to 31 denote parameter determination circuits which count the frequency of occurrence of the parameters 26 to 28 and output output values 32 to 34.

 Next, the operation of the parameter determination circuit will be described.

Address data 1 (i, j) is a matrix Input to the calculation circuit 2 and the matrix of equation (3) Is output. The address data 1 is input to the motion vector designating circuit 4. The motion vector designating circuit 4 outputs only the motion vector 7 of the block determined by the address data 1 among the motion vectors 6 of all the blocks supplied from the motion vector memory 5. In the product-sum operation circuit 8, the matrix -Sum operation of the motion vector 7 and Is performed), and a product-sum operation value 9 is output.

On the other hand, for the address data 11 (−i, j), Matrix determined by calculation circuit 12 and address data 11 A motion vector designating circuit 14, a motion vector memory 5, a motion vector 6 of all blocks, a motion vector 17 of a block determined by address data 11, and a product-sum operation value 1 using a product-sum operation circuit 18.
9 is output. The product-sum operation value 9 and the product-sum operation value 19 are the adder 1
Addition is performed with 0, and an added value 20 is output.

On the other hand, address data 1 and address data 11 It is input to the designation circuit 21. queue In the designation circuit 21, the matrix Matrix supplied from memory 22 Matrix determined by address data 1 and address data 11 of all patterns 23 Only output. Matrix Since all the patterns 23 do not depend on the type of image, they are obtained in advance offline.

Then add value 20 and matrix Is input to the product-sum operation circuit 25. The product-sum operation circuit 25 performs a product-sum operation, and obtains a parameter 2 representing the enlargement / reduction of the image
6, a parameter 27 representing the horizontal translation of the image,
A parameter 28 representing the vertical translation of the image is output. Each parameter 26, 27, 28 is a parameter decision circuit 2.
One is added to a counter which is input to 9, 30, 31 and counts the frequency of occurrence of parameters 26, 27, 28, respectively.

The above operations are performed for i = 1, 2,..., M−1, m, j = −n, −n +
After going for 1, ..., n−1, n, address 1 is changed to (i,
j), address data 11 is (−i, j), i = −m, −m
Similar operations are performed for +1, ..., m-1, m, j = 1,2, ..., n-1, n.

When the above processing is completed, the parameter determination circuit 2
At 9,30,31, the values with the highest number of appearances were 32,3
3, 34 are output as parameters of the entire image.

(Effect of the Invention) As described above, according to the present invention, even when a special motion other than the image enlargement / reduction and translation is included in the image, the image enlargement / reduction and translation parameters can be changed. It can be obtained with high accuracy only by simple product-sum operation and counting of occurrence frequency.

[Brief description of the drawings]

FIG. 1 is a diagram showing an example of how to give blocks and coordinates in an image according to the method of the present invention, FIG. 2 is a graph showing an example of a frequency distribution of parameter estimation values in each element, and FIG. 3 implements the method of the present invention. FIG. 4 is a block diagram of a parameter determining circuit according to an embodiment for the purpose of the present invention, and FIG. 4 is a diagram showing an example of a conventional method of giving blocks and coordinates in an image. 1,11 …… address data, 4, 14: a motion vector designating circuit, 5: a motion vector memory, 6: a motion vector of all blocks, 7, 17 ... a motion vector of a block determined by address data 1, 11, 8, 1
8,25: product-sum operation circuit, 9,19: product-sum operation value, 10: adder, 20: addition value, 26 ... Enlargement / reduction parameters calculated from blocks and motion vectors determined by address data 1 and 11, 27 ...
A horizontal translation parameter calculated from the motion vector of the block determined by the address data 1 and 11, 28
..... parallel movement parameters in the vertical direction calculated from the motion vectors of the blocks determined by the address data 1 and 11, 29.about.31... A parameter determination circuit, 32... Enlargement / reduction parameters of the entire image, 33... , The horizontal translation parameter in the horizontal direction, 34... The vertical translation parameter of the entire image.

Claims (2)

(57) [Claims] 1. Three types of parameters expressing the enlargement / reduction and translation of an entire image are determined simultaneously without convergence calculation from local motion vectors in block units obtained for two digital image frames. In the parameter determination method, a process of estimating a parameter using a part of the plurality of motion vectors among the plurality of block-based motion vectors is performed a plurality of times while changing a motion vector used for the estimation of the parameter. Obtaining the estimated values of a plurality of parameters as a whole, determining the most frequent estimated value among the obtained estimated values of the plurality of parameters as the parameters of the image, How to determine translation parameters. 2. A method according to claim 1, wherein the motion vector used for estimating the parameters is a motion vector of a block symmetric with respect to the center of the image in the horizontal or vertical direction. Method for determining the enlargement / reduction and parallel movement parameters of the image.