JPS63116193A - Affine transformation system for image - Google Patents

Abstract

(57) [Abstract] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

Detailed Description of the Invention (Technical Field to which the Invention Pertains) The present invention relates to a display device that stores and displays digital image data.
This relates to an affine transformation method for images that performs parallel translation, etc.

(Prior Art) In general, in affine transformation that performs scaling, rotation, etc. of image data, in order to determine each pixel data of the transformed image, a transformation matrix that defines the affine transformation is used to determine the corresponding pixel data on the original image for each transformed pixel. Since it is necessary to calculate the coordinate values of each pixel, there is a problem in that the conversion processing time for the entire image becomes long.

Conventionally, in response to such problems, several speed-up methods have been proposed that take advantage of the special characteristics of transformation for basic transformations such as scaling and one rotation among affine transformations. .

For example, a method of omitting coordinate transformation for each pixel by sequentially adding the scaling ratio or the variation rate in the vertical and horizontal directions due to rotation to the coordinate data of the original image along each coordinate axis (Reference 1), grid coordinate Focusing on the periodicity of , a method of accelerating image scaling by storing the values necessary for conversion in advance in a table and sequentially referencing this (Reference 2), converting rotation into two oblique axes This is a method (Reference 3) that decomposes and performs this oblique axis conversion at high speed by referring to numerical values necessary for conversion stored in a table in advance.

However, in these methods, for arbitrary affine transformations that combine scaling, rotation, and translation, it is necessary to decompose the given transformation into several basic transformations and perform multi-step transformation processing on the original image. There is a problem that it must be done.

In addition, all of these methods assume that the shape of the original image to be transformed is a rectangle parallel to the coordinate axes, but since the shape of the image after affine transformation is generally a parallelogram, this is not possible. When performing another affine transformation, or when considering cases where the image area to be replaced is a rectangle tilted with respect to the coordinate axis, there is a problem that the shape of the original image to be processed is restricted. .

[Reference 1] Kawakami et al. "Processing method of image processor MN8614" Information Processing Society of Japan Computer Architecture Study Group Materials, November 51-4, 1982 [Reference 2]
Tabata et al. “High-speed image scaling method using periodicity of lattice coordinates” Transactions of the Information Processing Society of Japan, 1983 Vol. 24 No. 5 [Reference 3] Tabata et al. “High-speed processing of image rotation using raster scanning and table reference” ” Journal of the Institute of Electronics and Communication Engineers, 1986 Vol. The purpose of this invention is to provide a means for processing arbitrary affine transformations at once and at high speed.

(Structure of the Invention) (Characteristics of the Invention and Differences from the Prior Art) In affine transformation of digital image data, the present invention provides affine transformation of the coordinate values of three points of a parallelogram that defines an original image, and of the three points. Define the output image obtained by
From the coordinate values of the point, for each pixel on the line segment with the first point as the starting point and the second point as the ending point in the output image, calculate the relative coordinate value of that output pixel with respect to the output pixel one position closer to the starting point, and one Calculates and stores the relative coordinate values of the original pixel to which the output pixel near the starting point corresponds to the original pixel to which the output pixel corresponds, and also calculates and stores the relative coordinate values of the original pixel to which the output pixel near the starting point corresponds, and For each pixel, calculate and store the relative coordinate value of the output pixel with respect to the output pixel one position closer to the starting point, and the relative coordinate value of the original pixel to which the output pixel corresponds with respect to the original pixel to which the output pixel one position closer to the starting point corresponds. By doing this, both the original image and the output image are divided into a line segment connecting the first point and the second point.・A line parallel to the line segment connecting the second point and the third point starting from each pixel above. It is characterized in that the image data is treated as a set of segment images, and the affine transformation of the image data is processed by repeating the affine transformation of line segment images as shown below.

That is, in outputting an affine transformed image, each pixel on the line segment connecting the first point and the second point of the output surface image and the original image is used as the starting point, and the pixels parallel to the line segment connecting the second point and the third point are For each pixel on a line segment, the process of setting the image data of the original pixel corresponding to the output pixel as the output pixel is repeated, and the coordinate values of the output pixel and original pixel on the line segment to be processed are , starting from the starting point in the direction from the second point to the third point, and sequentially adding the relative coordinates of the next pixel to the coordinate values of the pixel currently being processed, and the starting point coordinates of each line segment to be processed. is obtained by sequentially adding the relative coordinate value of the starting point of the next line segment to the starting point coordinate of the line segment currently being processed in the direction from the first point to the second point.

(Example) Hereinafter, an example of the present invention will be described using the drawings.

FIG. 1 is a diagram for explaining one embodiment of the present invention.

In the figure, reference numeral 1 denotes an original image definition point coordinate register that stores the coordinate values of three points that define the original image.

2 is an output image definition point coordinate register that stores the coordinate values of three points that define the output image.

3 is an output line segment pixel number calculation processing unit that calculates the number of pixels of the output line segment image excluding the starting point based on the coordinate data stored in the output image definition point coordinate register 2;

4 is a relative coordinate calculation processing unit that calculates relative coordinate values, and 5 is a control unit that controls affine transformation processing.

8.6 is the relative coordinate value of the output pixel with respect to the output pixel that is one position closer to the start point, and the relative coordinate value of that output pixel that is one position closer to the start point, for each pixel on the line segment whose start point is the first point and whose end point is the second point of the output image. This register stores the relative coordinate values of the original pixel to which the output pixel corresponds with respect to the original pixel to which the output pixel corresponds.

Reference numeral 6 indicates an original line segment image 12 relative coordinate table register, and reference numeral 8 indicates an output line segment image 12 relative coordinate table register.

9.7 is the relative coordinate value of the output pixel with respect to the output pixel one position closer to the starting point, and the relative coordinate value of that output pixel with respect to the output pixel one position closer to the starting point, for each pixel on the line segment with the second point as the starting point and the third point as the ending point of the output image. This is a register that stores relative coordinate values of the original pixel to which the output pixel corresponds with respect to the original pixel to which the output pixel corresponds.

7 is an original line segment image 23 relative coordinate table register, and 9 is an output line segment image 23 relative coordinate table register.

Figure 2 is a diagram for explaining the operation of Figure 1, and P
I, P2. P3 is the three points that define the original image, Ql,
Q2 and Q3 are three points that define the output image.

Figures 3 (A) and (B) show registers that store the relative coordinates in Figure 1 during the affine transformation process of the image data shown in Figure 2, and original line segment image 12 relative coordinate table register 6° original line segment. Image 23 Relative coordinate table register 7° Output line segment image 12
The specific values of the relative coordinate table register 8 and the output line segment image 23 relative coordinate table register 9° are shown.

Now in FIG. 2, Pi, P2. The original image defined by the three points P3 is Ql, Q2. Consider the process of performing affine transformation on an output image determined by the three points Q3.

First, points PL, P2. which define the original image and output image given from the host device. P3. and Ql, Q2. The coordinate data of Q3 is transferred to the original image definition point coordinate register 1 and the output image definition point coordinate register 2 in FIG.
2.

And points Ql, Q2. The coordinate data of Q3 is sent from the output image definition point coordinate register 2 to the output line segment pixel number calculation processing section 3 via the data bus 23. The output line segment pixel number calculation processing unit 3 calculates points Ql, Q2. Output line segment images IQI・Q21 and IO2・Q31 from the coordinate data of Q3
Number of pixels excluding the upper starting point ΔQ12. Find ΔQ23.

Here, ΔQ12. To find ΔQ23, for example, point Ql
.

Q2. The coordinate data of Q3 is Ql(X, , Y, l),
Q2(X, , Y, ,).

Q3(X, , Y, 3) may be calculated using the following formula.

ΔQ12 = wax <Ix, t LxL
IYqz Yqzl) ・=-(1)ΔQ
23 = wax (IX13 XqzL IYq3
Yqzl) ・-・・(2) However, (1), (2
maxL ) in the formula represents the maximum value among the values in parentheses.

The output line segment pixel number calculation processing unit 3 that performs this process can be configured by a general-purpose processor or a combination of a difference device and a comparison/determination device.

Calculated ΔQ12. ΔQ23 is data bus 2 in Figure 1.
4.25, they are sent to the relative coordinate calculation section 4 and the control section 5, respectively.

Next, in the relative coordinate calculation unit 4, ΔQ12. ΔQ23 and data bus 26
．． Definition points Pi of the original image and output image read from 1 and 2 through 27, P2. P3. Ql, Q2.
From the coordinate data of Q3, output line segment image IQI・Q2
For each output pixel except the starting point on Calculate the relative coordinate values of the corresponding original pixel.

These relative coordinate data are stored in the output line segment image 12, relative coordinate table register 8, output line segment image 23, relative coordinate table register 9, and original line segment image 12, respectively, in FIG.
The relative coordinate table register 6 and the original line segment image 23 are set in the relative coordinate table register 7 via the data bus 28.

Here, in order to obtain these relative coordinate values, the following processing may be performed.

That is, the coordinates of the image definition point in Fig. 2 are PL(Xpx-Ypt), P2(XP2-YF3)-P
3 (XP3-YF3).

Ql(X, 1.'/, i), Q2(X, 2.Y,,)
, Q3(Xq3.Y,,), then the line segment IQI・Q
The coordinate values of the i-th output pixel from the starting point on 21 and the corresponding original pixel are (Xqtz(1)yYq 2(i)),
[XP□z(1)yYp工,(i)).

The coordinate values of the j-th output pixel from the starting point on line segments IQ2 and Q31 and the corresponding original pixel are CXqzx (j) rYqxa (j)) v (X
P23 (j) r YF23 (j)), and since these can be found using the following formula, change i from 1 to ΔQ12 and j from 1 to ΔQ23, and calculate the difference between the values of i-1 and j-1. If you take them, you can find the relative coordinate values for each. However, sign (α) in equations (3) to (10) represents the sign of α, and [β] represents the maximum integer not exceeding β. FIG. 3 shows the respective relative coordinate values in the example of FIG. 2.

Next, the main part of the present invention is the table register 6°7.8.
．． Affine transformation processing of image data based on the relative coordinate values stored in 9 will be explained.

Here, 14 in FIG. 1 is an original image starting point coordinate register that stores the starting point coordinates of the line segment currently being processed on the original image from which image data is read, and 15 is the coordinate of the read pixel on the line segment. This is an original pixel coordinate register that stores .

Further, 16 is an output image starting point coordinate register that stores the starting point coordinates of the output line segment image to which the read pixel data is to be written, and 17 is an output pixel coordinate register that stores the coordinate data of the pixel to be written. It is a register.

10, 11, 12, and 13 are adders for adding the relative coordinate data of table registers 6.7 and 8.9 to the current coordinate data to be processed on each line segment;
．． This is an address conversion circuit that converts the XY coordinate values of 17 into addresses on the image memory 20.

First, the control unit 5, which controls the entire process, sends a "first starting point coordinate load signal" onto the control line 29.

In synchronization with this signal, the data bus 31.
32, the coordinate data of PL and Ql in the example of FIG. 2 are loaded into the original image starting point coordinate register 14 and the output image starting point coordinate register 16.

The coordinate data of PL and Ql is transmitted from the original image starting point coordinate register 14 and the output image starting point coordinate register 16 through the data bus 33.34 in synchronization with the "starting point coordinate load signal" sent from the control unit 5 through the control line 30. Current original pixel coordinate register 15. The current output pixel coordinate register 17 is loaded.

Furthermore, this coordinate data is sent to the address conversion circuit 18.19 through the data bus 44.45 and converted into a read address and a write address on the image memory, and according to these addresses, the image take on the original image is changed to the pixel on the output image. will be forwarded to. (In the example of FIG. 2, the image data of point P1 is set to point Q1.) Next, the control unit 5 sends out a "next pixel processing instruction signal" on the control line 30, and synchronizes with this signal. The current original pixel coordinate data sent from the current original pixel coordinate register 15 through the data bus 35 and the relative coordinate data of the next processing target original pixel sent from the relative coordinate table register 7 through the data bus 36 are added. The coordinate values of the next original pixel to be processed calculated here are set in the current original pixel coordinate register 15 through the data bus 3.

Similarly, in synchronization with this "next pixel processing instruction signal",
The current output pixel coordinate data sent from the current output pixel coordinate register 17 via the data bus 38 and the relative coordinate data of the next output pixel to be processed sent from the relative coordinate table register 9 via the data bus 39 are added to an adder. 1
3, and the coordinate value of the next output pixel to be processed calculated here is set in the current output pixel coordinate register 17 via the data bus 40. Then, the coordinate data of the new pixel to be processed stored in the register 15.17 is sent to the address conversion circuit 18.19 through the data bus 44.46, and is converted into a read address and a write address on the image memory, respectively. Accordingly, the image data on the original image is transferred again to the pixels on the output image. (In the example of FIG. 2, the image data of point P' is set to point Q'.) In this way, the control unit 5 controls line segment I02. The "next pixel processing instruction signal" is sent to the control line 30 for the number of pixels excluding the starting point on Q31 (that is, 6023 times), thereby starting from the first point of the output image and starting from the second and third points. The image data of the corresponding original image is successively set to each output pixel on a line segment parallel to the line segment connecting the . (In the example in Figure 2, pixel data corresponding to the line segment connecting points P1 and P'' is set for each pixel on the line segment connecting points Q1 and Q''.The mouth part in Figure 2 ) Next, when the conversion processing of the line segment image in the first column is completed, the control unit 5 sends a "normal line segment image processing signal" onto the control line 29. Then, in synchronization with this signal, the original image starting point coordinate register 1
The current original image starting point coordinate data sent from 4 through the data bus 41 and the relative starting point coordinate data of the next processing target original line segment image sent from the relative coordinate table register 6 through the data bus 42 are added to the adder 10. is added by
The starting point coordinate data of the next processing target original line segment image calculated here is set in the original image starting point coordinate register 14 via the data bus 43.

Similarly, current output image start point coordinate data is sent from the output image start point coordinate register 16 via the data bus 44 in synchronization with this signal, and the next processing target is sent from the relative coordinate table register 8 via the data bus 45. The relative starting point coordinate data of the output line segment image is added by the adder 12, and the starting point coordinate data of the next output line segment image to be processed calculated here is set in the output image starting point coordinate register 16 through the data bus 46. . (In the example in Figure 2, point p
The coordinate data of jll, Q"1 is set at 14.16 in FIG. The ``start point coordinate load signal'' is sent once and the ``next pixel processing instruction signal'' is sent 4023 times, thereby setting the pixel data on the original image corresponding to the output line segment image of the second column.

(In the example of FIG. 2, the part) In this way, the control unit 5 sends a "normal line segment image processing signal" once to the control line 29 every time it processes a line segment image for one column. The affine transformation of the line segment image is performed by sending the ``PA start point coordinate load signal'' once and the ``next pixel processing instruction signal'' 4023 times on the line 30, and converts this into the number of pixels excluding the starting point on the line segment IQI and Q21. (that is, ΔQ12), the entire image data is subjected to affine transformation.

(Effects of the Invention) As explained above, in the present invention, the affine transformation process of an image can be performed by four integer addition processes for each pixel, so that the conventional matrix operation for each pixel, that is, Compared to a method that performs addition and multiplication of real numbers four times each, it can process faster.

In addition, by processing the affine transformation of the image by repeating the affine transformation of line segment images parallel to each side of the parallelogram that defines the image, we have compared it to conventional image scaling and rotation methods. This has the advantage that digital image data defined by any parallelogram can be processed at once for any affine transformation.

[Brief explanation of the drawing]

FIG. 1 is a diagram for explaining one embodiment of the present invention, and FIG. 2 is a diagram for explaining the operation of FIG. 1. FIG. 3 shows specific values of the registers storing the relative coordinates shown in FIG. 1 in the affine transformation processing of the image data shown in FIG. 1... Original image definition point coordinate register, 2... Output image definition point coordinate register, 3... Output line segment pixel number calculation processing section, 4... Relative coordinate calculation processing section, 5... Control Part, 6...Mm image 12 relative coordinate table register,
7... Original line segment image 23 relative coordinate table register, 8
...Output line segment image 12 relative coordinate table register, 9
...Output line segment image 23 relative coordinate table register, 1
0, 11.12.13... Adder, 14... Original image starting point coordinate register, 15... Original pixel coordinate register, 16... Output image starting point coordinate register, 17... Output pixel coordinate register . 18, 19... Address conversion circuit, 20... Image memory, PI, P2. P3...Three points that define the original image, Q
l, Q2. Q3: Three points that define the output image. Patent applicant: Nippon Telegraph and Telephone Corporation

Claims (1)

[Claims] In affine transformation of digital image data, the coordinate values of three points of a parallelogram that define the original image and the coordinate values of three points that define the output image obtained by affine transformation of those three points are A means of remembering. For each pixel on a line segment with the first point as the starting point and the second point as the ending point in the output image, the relative coordinate value of that output pixel with respect to the output pixel one position closer to the starting point and the output pixel one position closer to the starting point correspond. means for storing the relative coordinate values of an original pixel to which the output pixel corresponds to the original pixel to which the output pixel corresponds; means for storing the relative coordinate value of the output pixel with respect to the output pixel of , and the relative coordinate value of the original pixel to which the output pixel corresponds with respect to the original pixel to which the output pixel one position closer to the starting point corresponds; means for calculating the relative coordinate values of each output pixel and the corresponding original pixel based on the coordinate values of three points defining the original image and three points defining the output image; , for each pixel on a line segment parallel to the line segment connecting the second and third points, starting from each pixel on the line segment connecting the first and second points of the output image and the original image. , the process of setting the image data of the original pixel corresponding to the output pixel to the output pixel is repeated, and the coordinate values of the output pixel and the original pixel on the line segment to be processed are changed from the second point to the third point. The coordinates of the starting point of each line segment to be processed are determined by sequentially adding the relative coordinates of the next pixel to the coordinates of the pixel currently being processed, starting from the starting point in the direction toward which the line is being processed. An affine transformation method for images, characterized in that the relative coordinates of the starting point of the next line segment are sequentially added to the starting point coordinates of the line segment currently being processed in the direction toward which the image is directed.