JPS6097473A - Method and apparatus for rotation processing of picture - Google Patents

Abstract

PURPOSE:To obtain a rotation picture at high speed by making oblique axis conversion of original picture data of a picture, and then making processing of enlarging and reducing, and further making oblique axis conversion. CONSTITUTION:A digital picture is stored in a picture memory 40, and oblique axis conversion T1 is made on picture data by a microprocessor 10. Original picture data in a rectangular area of N words X M rows on a memory are transferred word by word in order of number, and the oblique axis is converted by address of destination of transfer. Data generated by this conversion are enlarged and reduced by an enlarging and reducing device 30 basing on converted row and column T2. Processing is made making magnification is horizontal direction alpha=costheta and making magnification of vertical direction beta=sectheta. A desired rotation picture is obtained by making oblique axis conversion T3 on enlarged and reduced picture data.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Application of the Invention] The present invention relates to an image rotation processing method, in particular, image 1#! The present invention provides high-speed processing means suitable for editing documents containing documents, verifying stamps, etc.

[Background of the invention]

Let us consider a process in which a digital image, defined as a set of pixels in a square grid, is rotated by a total angle of θ radians. Image rotation is a type of affine transformation and is defined by the following equation.

These are the pixel coordinates of the first transfer.

Conventional image rotation methods (for example,
61, ■4 Children Communication Society research and death fee IE78-12, ■C
computer, IEEE+ p, p, 24~
26゜Jun 61983 etc.) generally repeats the locus image transformation of equation (1) pixel by pixel, which makes high-speed processing difficult. On the other hand, the applicant, etc.
Faster tribute? For L objective, some adjacent lI! He invented an image rotation method (Japanese Patent Application No. 57-176151) that transfers all j elements at once. This method allows arbitrary rotation angles, but during processing +i depends on the rotation angle, and (2n-1)×π
There was a drawback that the processing time was maximum at a rotation of /4 radian (1≦n≦4). The applicant has invented a second image rotation method (Japanese Patent Application No. 58-89119) to solve this problem, but the rotation angle is (2n-1)xπ/4
I was dealing with the case of radians.

[Purpose of the invention]

An object of the present invention is to provide a means for realizing high-speed rotation of both edges at an arbitrary angle θ (-π≦θ≦π).

[Summary of the invention]

Then, Wu α) can be transformed as follows.

Busy here, matrix Tl + 'I'11 is oblique axis transformation T2
represent the transformation matrices of scaling, respectively. Therefore, as shown in Figure 1, the original image is subjected to oblique axis transformation (T+).
, scaling (T z ), and oblique axis transformation (Ts), a rotated image at an angle θ is obtained. Among these, the oblique axis transformation (T+ and Ts) is simple to process, and the block transfer method developed by the applicant and others (Information Processing Society of Japan Journal: Proposal of a memory address control method using two-dimensional block transfer and document Application to image processing: Showa 5
8-move-July) can be executed at high speed. Furthermore, for scaling (T2), a high-speed processing method invented by the applicant (% Application No. 57-71237) can be used. As described above, the present invention is characterized in that it speeds up the rotation process of the picture iron by combining oblique 1M conversion and scaling.

However, when θ→Sat π/2, 1tan θ1→■.

Since the 18 mark θ1→(3), the matrix T2゜1゛3 in equation (2) becomes undefinable. In addition, as 1tan θ1 and 1 vote θ1 increase, oblique axis transformation (Ts) and scaling (T2) gold height i? It becomes increasingly difficult to achieve this.

This problem can be solved by selecting the following combination of oblique axis lA'', conversion and enlargement + iris, depending on the rotation time 00ψ range.

a) −π/4≦θ×π/4.3π/4≦0<π.

When −π≦θ×3π/4 At this time, each element of the variable matrix has the following range.

0≦Giant anθ1≦1, 1/y'251 CO8θ
1≦1.1≦Hecθl≦V'2.

b) π/4≦, θ×3π/4. −3π/4≦θku−π/
In the case of 4, at this time, ) exchange? - Each element of the J' column has the following If area.

0≦1 starvation θI ≦1 + 1 /V7S l sinθ
1≦1, 1≦l cosec θ1≦v'z It is already well known that the technology to obtain a product 7'jj, which is scaled at a magnification of about 1/v/T-V/T, has been established ( For example, the 13th Painter Ichika Kouken Conference, p.
, 183-186, December 1982), the problem of lowering both pages can be solved by selecting a combination of transformation matrices according to the rotation angle θ (α).

[Embodiments of the invention]

Hereinafter, the present invention will be explained in detail by way of examples.

FIG. 2 shows an example of a hardware configuration for realizing the present invention. In the figure, 10 is a microprocessor, 20 is a main memory, 30 is a scaling device, and 40 is an image memory.

For the sake of explanation, the one digital painter targeted here is
2Ili! It is assumed that the image is an image (i.e., a collection of pixels g waiting for either white or black gray level), but the present invention deals with one multilevel image (i.e., a collection of pixels with multiple gray levels). can also be applied.

As shown in Figure 3, the image memory has a two-dimensional address in pixel units, and the W
Focus (1 word) stroke data 'lfr - Accessed with the extraction/insertion command. In the figure, the mark indicates the starting focus position of one word. The access mode is specified by the parameter t, and when t=o, the horizontal direction is accessed; when t=1, the vertical direction W bintra access is performed. A method for realizing such a two-dimensionally accessible image memory is well known and is described in other literature (for example, IEEg, Transo
n computersXc-27, A, 2.1)p
, 113-125. Feb., 1978) is also detailed, so the explanation will be omitted.

A microprocessor 1o controls the operation of each device, and stores programs and data necessary for this in a main memory 20. The processing procedure performed by the microprocessor 10 is shown in FIG.

If the transformation matrix is expressed (i.e. -π/4≦θ)π/4°3π/
The processing contents will be explained below, assuming that 4≦θ〈π, -π≦θ〉3π/4) is −lery (FIG. 5).

(1) Regarding process 100, perform oblique axis transformation TI on the original image data on the image memory.
administer. The detailed procedure is as shown in FIG. Processing 1
dt of the oblique axis transformation at 00 is shown in FIG. As shown in Figure 6, the original image data in a rectangular area of N words (width) and XM rows (height) on both edge memories is transferred one word (i.e., W focus) in the numerical order of the figure. It is something. Start address t of the transfer source (i.e. original image data): (
X+.

shall be.

a) Process 110: Specify continuous W focus access in the horizontal direction by setting the parameter t indicating the access mode of the image memory to 0. The transfer source's X address (Xl
) and X address (y+), forwarding destination X address (xl)
Initializes the X address (V2) and the line counter (LC). That is, X('-XI.

Y2. Let LC←0.

b) Process 120: Increment the line counter (LC) t- and initialize the word counter (WC). That is, Lc←LC+1.
Let WC←0.

C) Process 130: Increment the word counter (WC). That is, WC4-WC+1. Kokote and WC are counters that indicate the number of transferred words in the relevant row of image data.

d) Process 140: Transfer one word of image data from address (xl, y+) to address (Xg + )'2) in the image memory. However, the address of the image memory is indicated by the position of the first focus (marked with a square in FIG. 3) of the word.

Also, there is an access mode Idt=0-'C.

e) Process 150: When the word counter (WC) is equal to N, it means that all words of the corresponding row-hi of the image data have been transferred, so in this case proceed to process 160. For parts where wc is less than N, the process proceeds to step 170 to continue data transfer regarding the row.

f) Process 170: Transfer source X address (x+), m destination X address (x
t) is updated as shown below. That is, XI←Xz +W,
Let X2←x2+W. After the update, the process returns to step 130.

g) Process 160: When the line counter (LC) is equal to M, it means that all image data of N words and XM rows have been transferred, so in this case, process 100 is ended. If LC is less than M, proceed to process 180.

h) Process 180: Update the transfer source and transfer destination X addresses and X addresses as follows. That is, X1←x,,y.

← Y+ it Δ xl ← Δ Xz+tan θ
l X2 °X2+[ΔXzl), Yz←y2←1. However, the dself symbol [ ] represents a Gaussian symbol.

(2) Regarding process 200, convert the image data generated by process 100 into a transformation matrix T2
Scale based on. That is, the horizontal direction (X-axis direction)
Let 4 times 4 be α=cos θ, and let the magnification in the vertical direction (Y-axis direction) be β−cos θ. Enlargement/reduction processing is done by enlarging/reducing device +t1
30, and the processing results are stored in the image memory 40. For the enlarging/reducing device d30, a high-speed enlarging/reducing processing method invented by the applicant (Japanese Patent Application No. 71237/1982) can be used. In particular, this method is high-speed when the relational expression of multiplication factor constant/integer variable is satisfied. Now, when the integer constant in the above equation is 32 and the rotation angle is θ = π/4, consider as an example, the magnifications α and β are α-1/v'2?
32/45#0.711, β=V-d-32/23! =
This will be approximated as i1.391.

In this column, the error in magnification due to approximation is about 1 to 2 inches, which is considered to be within a practical tolerance range.

(3) Processing 300 As a result of processing 200, the image data that has been scaled up or down is subjected to oblique axis λ conversion T3. The procedure is shown in FIG. The contents of the oblique axis transformation in process 300 are as shown in FIG. As shown in Figure 8, image data within a rectangular/robust area of Q columns (width) x P words (height) on the image memory is transferred word by word (i.e., W bits) in the numerical order of the figure. It is something to do. However, a rectangular area of Q columns×P words is determined so as to include the image data after being enlarged or reduced by the process 200. Furthermore, access to the image memory is performed using W bits that are continuous in the vertical direction as one word.

The first address of the transfer source on the image memory t=(X+-Y
+), and the first address of the transfer destination k (X2, Y2
).

a) Process 310; Parameter t indicating the access mode of the image memory
By setting 1 to 1, access of continuous W focus in the vertical direction is specified. In addition, the X address of the transfer source (X
+) and X address (yt), ? Destination X address (x2
), the X address (y2), the column counter (CC), etc. are initialized. That is, xl←ma, *Yt'-Y1+x2"
E"'21 Yz"Zt + CC←O1Δy2←0
shall be.

b) Operation 320: Column counter (CC) '5 - increment and word
Initialize the counter (WC). That is, CC←CC+1
．． Let WC←0.

C) Process 330: Increment word counter (WC) t-. That is, WC4-VVC+1. Here, WC is picture 1
3J! This is a counter indicating the number of transferred words in the relevant column of data.

d) Process 4340: Transfer one word of image data from address (X+*)'I) to address (X21 Yz) on the image memory. However, the address of the image memory is indicated by the position of the first bin of the word (marked with a square in Figure 83). Also, the access mode is t=1.

e) Process 350: When the word counter twc> is equal to P, it means that all words on the corresponding column of stroke data have been transferred, so in this case, proceed to process 360. If WC is less than 2, the process advances to step 370 and continues data transfer regarding the column.

f) Process 3702 Transfer source X address (yI), transfer destination X address (y
2) is updated as follows. That is, yl←y 1 +
Let W*Yz←Yz+W. After updating, process 330
Return to

g) Process 360: When the column counter (CC) is equal to Q, it means that all the image data of Q columns x P words has been transferred, so in this case,
Process 300 ends. If CC is less than Q, process 3
Proceed to 80.

h) Process 380: Update the transfer source and transfer destination X addresses and X addresses as follows. That is, x1←x1-4-1゜Y t''-
Y t + X t←Xt+1. Δy2←ΔY z
+ tanθ.

Let y2←Y2+[Δy2]. However, the symbol [ ] represents a Gauss symbol.

The above processing procedure is performed when the transformation matrix is expressed as TI= (i.e. -π/4≦θ<π/4°3π/4≦θ<<π,
This relates to image rotation in the case of −π≦θ (−3π/4). In addition, when the transformation matrix is expressed as
(In the case of π/4 ≦ θ - π/4) As shown by the Vi transformation matrix TI, it is sufficient to add rotation by an angle π/2 to the oblique axis transformation, and the process is self-evident, so the explanation will be omitted. .

In the above embodiment, a simple processing method was introduced for convenience of explanation, but it is easily understood that various processing methods such as those described in f are also possible.

■ In processes 130, 140, 150, and 170 in FIG. 5, the program processing of the microprocessor 1o
We have shown a method of transferring data word by word on the image memory, but in order to speed up this process, DM A (1) i
Rect Memory Access) ic method can also be used. DMA allows high speed data transfer without microprocessor 10 intervention. D
Since the MA method is well known, its explanation will be omitted. Similarly, the processes 330, 340, 350, and 370 in FIG. 7 can also be replaced with DMA transfer.

(2) In FIG. 4, a method for realizing image rotation in three steps of processing 100, 200, and 300 is shown. As an alternative method, processes 100 and 2001. Rirui processes 200 and 30
0 can also be processed in parallel.

(2) Although the above embodiment deals with binary images, the principles of the present invention can also be applied to multivalued images and color images.

■ In this example, when selecting a combination of oblique axis transformation and scaling according to the range of rotation angle θ, the area of θ is equally divided by an integral multiple of the angle π/4. θ in the way
The area may be divided.

〔Effect of the invention〕

According to the present invention, the following effects can be obtained.

(1) The present invention is characterized in that image rotation is realized by a combination of oblique axis conversion and scaling. Oblique axis transformation is simple to process, can process several pixels at once, and
High-speed data transfer using DMA can also be used. For scaling, a high-speed processing method invented by the applicant can be applied. Therefore, in comparison with the conventional method in which the one-vote conversion for rotation is repeated pixel by pixel, according to the present invention, it is possible to realize image rotation at high speed.

(2) In the image rotation method using block transfer, which the applicant has already invented (Japanese Patent Application No. 176151/1982), the processing time depends on the rotation angle, and the rotation of (21m-1)Xπ/4 radians (l≦n≦ 4) had the disadvantage that the processing time was at its maximum, but the present invention eliminates this disadvantage.

[Brief explanation of the drawing]

FIG. 1 is a diagram explaining the principle of the present invention, FIG. 2 is a hardware configuration diagram of an embodiment of the present invention, FIG. 3 is a word configuration diagram of an image memory of the present invention, and FIG. 4 is a diagram of the present invention. Flowchart showing the overall processing flow of the invention, FIG. 5 is a processing flow diagram of oblique axis transformation TII in the invention, FIG. 6 is a data arrangement diagram in oblique axis transformation T of the invention, and FIG. 7 is a processing flow diagram of the oblique axis transformation '113 in the present invention, and FIG. 8 is a data arrangement diagram in the oblique axis transformation T3 of the present invention. 10...Microprocessor, 20...Main memory, No. 1, No. 2, No. 4, No. 5F+ No. b, Fig. 0

Claims (1)

[Claims] 1. A first step of converting the orthogonal coordinates -11+K into oblique axes forming an intersecting angle θ, and (g) θ or sin θ or an approximation thereof in the direction of one axis of the orthogonal coordinates. A method for image rotation processing, comprising a second step of reducing the image by a magnification and enlarging it in the other axis direction by the maximum θ or cosec θ, or an approximate magnification thereof. 2. Changing the oblique axis conversion formula in the first step and the expansion/reduction formula in the second step according to the range of the intersection angle 0.
1. A method for rotating an image as described in Section 1 of Section 1. 3. Memory" 2nd picture 1) It has means for performing oblique axis transformation of the image and means for measuring the enlargement/reduction of the transformed image in order to transfer each block of consecutive memory addresses. and t-featured image rotation processing device.