JPH0668245A - Image rotation correction processing method and image rotation correction processor - Google Patents

Abstract

PURPOSE:To provide a rotation correction device rotating an original image in an optimum direction by performing the rotation correction for an image inputted in its inclined state. CONSTITUTION:When the inclination of an image is detected in an inclination detection circuit 40, the rotation correction is realized and a corrected image is outputted by imparting proper rotation correction amount to a rotation correction circuit 50 by the inclination detection circuit 40 and a control input device 70. Therefore, because the rotation of the image is decomposed into the parallel moving in horizontal/vertical directions and the rotation correction is realized by only a bit shift and a logical calculation, an effect capable of remarkably improving processing speed is given.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention is to display an image read by an image input device such as a scanner on a display, and as a preprocessing for character recognition and the like.
The present invention relates to an image rotation processing method and apparatus for rotating a document and an image in arbitrary directions.

[0002]

2. Description of the Related Art When a document or an image read by an input device such as a scanner is displayed on a display, the image is displayed sideways or downward due to restrictions such as the standard of the input device and the size of the paper surface of the original image. There is a case.
Further, when an image is input by a handy scanner or an image after copying is input, the image may be read with an inclination. When using such an image for soft facsimile or desktop publishing, it is necessary to perform rotation correction in the direction that is most readable. Further, in the case of realizing character recognition of a document image or the like, it is expected that the recognition rate can be improved by using the rotation correction of the document image as preprocessing. Therefore, image rotation correction is a very important process. Conventionally, two-dimensional affine transformation is famous as an image rotation method. As shown in FIG. 2A, this method is a two-dimensional linear transformation in which a figure is translated and a figure is rotated by an angle θ. Here, for simplicity, rectangles P ₁ , P ₂ ,
It is shown that P ₃ and P ₄ are translated to the coordinate origin by l and m in the X and Y directions, respectively, and are rotated about the origin by an angle θ. The original coordinates P (x, y) are converted to the coordinates Q in the equations (1) and (2)
The transformation matrix T _PR of the two-dimensional affine transformation is shown in (X ″, Y ″).

[0003]

[Equation 1]

[0004]

[Equation 2]

By the way, the number of pixels of an A4 size binary image is 1728 horizontal × 2376 vertical dots (200 dpi), and when the above-mentioned Affine conversion is applied to each of these pixels, the processing amount is enormous. Become. Floating point arithmetic is required because it includes sin θ and cos θ arithmetic. Therefore, a high-speed CPU is required to realize it with software, and a complicated configuration is required to realize it with hardware, which is not realistic.

The present invention has been made to solve the above-mentioned problems, and realizes image inclination correction at high speed,
Another object of the present invention is to obtain an image rotation correction processing method and apparatus that can reduce the load on the CPU.

[0007]

When an input image read by an image scanner or the like is displayed on the display with a tilt, or when tilt correction is performed as preprocessing in the recognition processing of a document image, an arbitrary angle θ is set. It is necessary to rotate the image at high speed. The number of pixels constituting one A4 sheet is about 500 kbytes, and the conventional image rotation correction device processes each bit, which results in an extremely large amount of calculation and a problem that the processing time increases. Was there. Further, faithfully executing the two-dimensional Affine transformation requires floating point arithmetic, which makes it difficult to process the image rotation correction at high speed. Therefore, an object of the present invention is to realize a rotation correction processing method and apparatus capable of processing each image in units of bytes instead of processing in units of bits and further processing only by bit shift and logical operation of bits.

[0008]

According to an image rotation correction apparatus of the present invention, when an input image is input with an inclination, an original image read in an image rotation correction processing method for correcting the inclination according to the amount of inclination is used. Binarizing means for binarizing, image output means for displaying a binarized image, inclination detecting means for detecting the amount of inclination of the read image, rotation correcting means for performing inclination correction by Affine transformation, An image rotation correction processing method comprising: a control input unit for inputting a tilt correction amount from an external input device; and a control unit for controlling the tilt detection unit and the rotation correction unit.

The rotation correction circuit processes the rotation correction in byte units by executing the rotation correction by the parallel movement of the X axis and the parallel movement of the Y axis. Therefore, the bit shift and the logic for each pixel of the input image are performed. The feature is that the affine transformation is executed only by calculation to correct the inclination.

[0010]

According to the present invention, when the image is input with a tilt, the tilt amount is determined and detected, and the rotation is corrected according to the detected correction amount. Since the inclination correction means of the present invention corrects the rotation amount for each pixel of the input image only by bit shift and logical operation, the operation amount can be greatly reduced. In addition, since the arithmetic processing can be performed in byte units, high speed arithmetic can be performed.

[0011]

FIG. 1 is a block diagram showing the overall configuration of an image rotation correction device according to an embodiment of the present invention.

Image data read by a data input device such as an image scanner is binarized by a binarization circuit 10 and then displayed by an image output device 20 such as a display. , 30 image memories. The tilt detection circuit 40 detects the tilt of the image and outputs the tilt amount. As the configuration of the inclination detection circuit 40, particularly for a document image, a method of utilizing image area division or scanning the entire document from top to bottom in an arbitrary angular direction to determine the number of inversions of black and white pixel values on a scanning line is performed. A method of counting and statistically processing the inclination of a document image (Ishitani, "One method for detecting the inclination of a document image", 1
Proposals have been made such as the Spring 1992 Faculty of Science, D-570).
Further, the rotation input amount of the image can be arbitrarily input by the control input device 70. The rotation correction circuit 50 rotates the image by inputting the correction amounts obtained by the inclination detection circuit 40 and the control input device 70 to the rotation correction circuit 50.
The new target image whose rotation has been corrected is output again to the image memory 30 and the image output device 20. Furthermore, the tilt detection circuit 4
At 0, the tilt amount θ is calculated again, and the tilt amount is determined by the tilt allowable amount ep
When epc> θ for c, the rotation correction process ends.
Here, the allowable tilt amount epc depends on the entire system configuration, but it is desirable that epc <0.5 [deg].

Next, each of the above components will be described in detail below.

The binarization circuit 10 binarizes a gradation-displayed natural image or a color still image by performing threshold processing according to the gradation or luminance value of the image. When the input image data is a binary image, the binarization circuit 10 can be eliminated.

The inclination detection circuit 40 detects the inclination of the input image from the original image. In particular, in the case of character recognition processing of a document image, a method of detecting the amount of inclination by utilizing the statistical property of input data has been proposed. Although the inclination detection circuit 40 can use these inclination detection methods, when the image data is handled by desktop publishing or the like, or when the image is displayed on the display by soft facsimile, the inclination of the image is detected. Can be visually confirmed on the display. Therefore, the image tilt amount automatic detection means is not always necessary. Therefore, the inclination correction of the image is easily realized by the following method.

When displaying an image on a display such as a liquid crystal display, for example, 640 × 40 in a personal computer.
Each display has a specific resolution like 0 dots. Normally, the resolution of the display is lower than that of an image input device such as a scanner. Therefore, when the image rotation correction is performed, it is impossible to rotate the image with arbitrary accuracy due to the restriction of the dot pitch. FIG. 3 shows the relationship between the number of dots and the inclination of the image. The rotation frequency is a value of k indicating the angle θ as shown in FIG. The rotation angle θ decreases as k increases. The relationship between the rotation frequency and tan θ, θ, cos θ is shown in the table. 1/256 = in image rotation correction
If it is 0.22 [deg], it is considered that the rotation accuracy is sufficient. Further, in the tilt detection circuit 40, an error of about 0.2 [deg] on average occurs in the tilt automatic detection method using the statistical method. Even for human visual characteristics, such an error is considered to be within an allowable range.

In an apparatus for processing an image on a display such as a soft facsimile, a user operates a control input section 70 through a man-machine interface to input a rotation control signal to realize a rotation correction. It is common. When a keyboard, a mouse, a pen input tablet, or the like is used as the control input unit 70, an arbitrary angle θ in FIG. 3 can be selected and the rotation control amount can be input by the control input unit 70. The display image can be rotated at an arbitrary angle depending on how to select the angle θ. A method of selecting the angle θ will be described according to Example 1.

Example 1 When the display image is inclined 15.4 [deg] clockwise with respect to the original image, the angle θ for correcting this is selected.

(Positive and counterclockwise inclinations indicate the amounts of inclination deviations in the clockwise and counterclockwise directions.) Correction procedure [I] Number of corrections 1st time 2nd time 3rd time Rotational frequency k = 2 k = 5 k = 7 Rotation direction Counterclockwise Counterclockwise angle Angle θ [deg] 14.0 1.8 0.45 Tilt [deg] Positive 1.4 Anti 0.4 Positive 0.05 Correction procedure [II] Number of corrections 1st 2nd 3rd rotation k = 2 k = 6 k = 7 rotations Direction Counterclockwise Counterclockwise Counterclockwise Angle θ [deg] 14.0 0.89 0.45 Tilt [deg] Positive 1.4 Positive 0.51 Positive 0.06 Correction procedure [III] Number of corrections 1st 2nd 3rd 4th 5th 6th rotation k = 4 k = 4 k = 4 k = 4 k = 7 k = 7 Rotation direction Counterclockwise Counterclockwise Counterclockwise Counterclockwise Counterclockwise Counterclockwise Angle θ [deg] 3.6 3.6 3.6 3.6 0.45 0.45 Tilt [deg] Positive 11.8 Positive 8.2 Positive 4.6 Positive 1.0 Positive 0.55 Positive 0.10 As shown in the above correction procedure, rotation of an arbitrary angle can be realized by selecting the rotation frequency k. Furthermore, the correction procedure [II
In [I], the rotation frequency can be fixed to k = 4 or 7, and only one of them can be selected to realize rotation at an arbitrary angle. In this case, the correction error is within 0.22 [deg]. As shown in this example, even when the rotation direction and the rotation amount are fixed to some extent, it is possible to rotate an arbitrary amount, and when considering the man-machine interface, it is advantageous that the selection items can be reduced.

In the rotation correction circuit 50 shown in FIG. 1, the following policy is set in order to perform rotation correction easily and at high speed.

1. Floating point arithmetic is not performed to reduce the calculation processing.

2. Even if the accurate moving position is calculated, it cannot move to the calculated position due to the limitation of the number of pixels of the display. Therefore, it is not necessary to accurately calculate the moving position.

(Regarding the rotation accuracy, θ = 0.25 [d
eg] is sufficient) 3. Rotation can be achieved by two parallel translations in the vertical and horizontal directions.

4. In order to avoid floating point arithmetic, the transformed image cannot be perfectly projected onto the original image.

As a method of realizing the rotation correction circuit 50, the rotation correction circuit of the present invention will be described in detail. First, the affine transformation that rotates about the origin O is shown in equation (3).

[0026]

[Equation 3]

This relationship is shown in FIG. When the formula (3) is modified, it becomes as follows.

[0028]

[Equation 4]

[0029]

[Equation 5]

U and v are defined as the amounts of movement in the X and Y directions.

[0031]

[Equation 6]

[0032]

[Equation 7]

Here, from the table of FIG. 3, when the rotation frequency k is 3 or more, cos θ≈1, and therefore equations (6) and (7) are as follows.

[0034]

[Equation 8]

[0035]

[Equation 9]

However, INT (X) is a function for converting a decimal value to a real value (truncated below the decimal point). If the amount of tilt correction of an image is not so large, cos θ can be approximated to 1 and therefore is shown in equations (8) and (9). Can be transformed into Since tan θ = 2- ^k , the movement amounts u and v can be expressed only by bit shift.

As described in Policy 3, the rotation is in the horizontal direction,
It can be expressed by two parallel translations in the vertical direction. The position of the original pixel, the position after the vertical movement, and the position after the horizontal movement are Q (X _j , Y _j ), Q '(X' _j , Y ' _j ), and Q "(X" _j , Y ", respectively. _j ).

[0038]

[Equation 10]

Q ″ is a converted pixel value that is rotated and moved. Here, u (Y) is a function of the position Q (X, Y) of the original pixel. This is a function of Y ′ after moving v. The following conversion is performed to

[0040]

[Equation 11]

FIGS. 4 and 5 show the above-described policy and the image rotation procedure using this simple rotation correction means.

FIG. 4A shows an original image. Θ = -1
It shows the rotation of 4.4 [deg] and tan θ = -1/4. FIG. 4B shows the vertical translation. With coordinates P (X _o , Y _o ) as the origin,
tan θ = -1 / 4 (the code depends on the value of tan θ depending on the position of the target pixel) and is translated in the vertical direction.

[0043]

[Equation 12]

Depending on the X position of the original image, it moves by v.

The horizontal translation is shown in FIG. 4 (c). By the same procedure as in the vertical direction, equations (8) and (9)
(10) In accordance with (11), the object is moved in parallel to the coordinate Q ″ (Xj ″, Yj ″).

[0046]

[Equation 13]

FIGS. 5A, 5B and 5C show the state of rotation correction when θ = −26.6 [deg] and tan θ = −1 / 2. Formula (12)
Using (13), the rotation can be corrected in the same manner as in the case of the rotation frequency k = 2. At this time, the following expressions are used for the movement amounts u and v.

[0048]

[Equation 14]

By performing image rotation correction by the method shown in FIGS. 4 and 5, rotation correction with a small amount of calculation can be realized. 6 and 7 show in detail how the rotation correction device according to the present invention is implemented by hardware or software.
Here, it is assumed that the original image is rotated clockwise with θ = 3.6 [deg] rotation degree k = 3.

FIG. 6A shows an example of elements of each register on the memory. This corresponds to each pixel value on the output means (display) bit by bit. Figure 6
The vertical translation is shown in (b). When the original image is stored in the two-dimensional array A (i, j) (array A is 1 byte), in parallel translation, rotation of tan θ = -1/8 is 8 bits as shown in the figure. Since each memory element (array) A is replaced by a memory element (array) B, it can be realized. The conversion relationship in the case of rotation correction around A (i, j) is shown.

[0051]

[Equation 15]

FIG. 6B shows an example of the case where the rotation correction is performed around A (i, 1). When this is implemented by software, it is only necessary to change the address on the memory, so processing can be performed at extremely high speed. Also, the rotation frequency k is k
When <3, processing cannot be performed in byte units, but it can be handled in the same way by performing bit mask processing.

Next, horizontal translation will be described.
Since the rotation is tan θ = -1 / 8, processing is performed every 8 bits, and each array is collectively processed in every 8 bits in the vertical direction. For simplification, the array B is changed to B (8n + l, j) [l = 0,1,2, ...,
7, n = 0, 1, 2, ...], which is a block B _nj . When rotating around the point P, each block B _nj can be moved in parallel in the horizontal direction by bit shift. The pixel block after conversion is C _nj . For bit shift of 8 bits or more, it is not necessary to prepare a shift register of 8 bits or more by exchanging storage elements and using bit shift together. As shown in Expression (11), the horizontal movement requires a correction term tan ² θ depending on the horizontal position, but this operation also determines the rotation frequency k (angle θ), and thus the X coordinate Since it can be uniquely determined only by the position of, it can be executed only by the bit shift process. Also, for bits that overflow due to bit shifting, it is necessary to take them into a buffer and insert them into blocks before and after. Even when the pixel value is not determined by the correction in the X direction, that is, when the pixel is missing, it is possible to prevent the pixel missing by replacing the preceding and following pixel values. As described above, it has been shown that the rotation correction around the point P can be realized only by bit shift and replacement of array elements. 6 and 7 show an example in which the rotation frequency k = 3, the same operation can be realized for other rotation frequencies. While image processing such as Affine conversion is performed on a bit-by-bit basis, the image processing according to the present invention has an advantage that it can be processed on a byte-by-byte basis.

Image processing performed on the display may require 90-degree rotation correction when the original image is read in the horizontal direction. Therefore, a means for realizing the 90-degree rotation correction will be described below. FIG. 8A shows a model when a horizontally oriented image is vertically corrected. A _ij is a 1-byte original pixel register. Rotation correction of 90 degrees can be realized by fetching each of these registers every m bits by the shift register and the m-bit buffer shown in FIG. 8B and moving them to the position of the affine conversion register C _ji . FIG. 8B shows an example where l = 8 and m = 8. Rotational corrections of 180 degrees and 270 degrees can be realized by repeating 90-degree rotation corrections twice and three times.

In the above embodiment, tan θ = 2 is used for correcting the inclination of the image.
It is shown that arbitrary rotation correction can be realized by arbitrarily setting k using the rotation frequency k of ^-k , and an image rotation correction processing method capable of performing rotation correction processing in byte units as a method of embodying the rotation correction means in FIG. The device has been described. As a result, high-speed rotation correction can be realized with a simple structure.

[0056]

As described above, according to the present invention, since the rotation of the image is decomposed into the parallel movement in the horizontal and vertical directions, the rotation correction is realized only by the bit shift and the logical operation, so that the processing speed can be greatly improved. Have an effect. Furthermore, since it is a simple algorithm, it can be easily implemented as software and hardware.

[Brief description of drawings]

FIG. 1 is a block diagram showing an overall configuration of a rotation correction processing method and apparatus according to an embodiment of the present invention.

FIG. 2 is a diagram defining a two-dimensional affine transformation and movement amounts u and v.

FIG. 3 is a diagram showing a relationship between a definition of a rotation frequency and a rotation angle.

FIG. 4 is a diagram for explaining image rotation correction means (rotational frequency k = 2).

FIG. 5 is a diagram for explaining image rotation correction means (rotational frequency k = 1).

FIG. 6 is a diagram showing a specific example for realizing image rotation correction means (Y direction parallel movement).

FIG. 7 is a diagram showing a specific example for realizing image rotation correction means (translation in the X direction).

FIG. 8 is a diagram illustrating rotation correction means when the rotation angle is 90 degrees.

[Explanation of symbols]

 10 ... Binarization circuit 20 ... Image output device 30 ... Image memory 40 ... Tilt detection circuit 50 ... Rotation correction circuit 60 ... Control unit 70 ... Control input device 71. ..Inclination detection control signal 72 ... Rotation correction control signal 73 ... Control input signal 74 ... Image data

Claims (3)

[Claims] 1. An image rotation correction processing method for correcting an inclination according to an inclination amount when an input image is input with an inclination, a binarizing unit for binarizing a read original image, and a binarization. Image output means for displaying the displayed image, tilt detection means for detecting the tilt amount of the read image, rotation correction means for performing tilt correction by affine transformation, and control input means for inputting the tilt correction amount from an external input device. And an image rotation correction processing method comprising: a tilt detection unit and a control unit that controls the rotation correction unit. 2. An image rotation correction processing apparatus, wherein image rotation correction processing means performs affine transformation by only bit shift and logical operation for each pixel of an input image to correct the inclination. 3. The rotation correction of the image rotation correction processing means is X.
3. The image rotation correction processing device according to claim 2, wherein the rotation correction is processed in byte units by executing the parallel movement of the axes and the parallel movement of the Y axes.