JPH04215193A - Picked-up image corrector - Google Patents

Abstract

PURPOSE:To improve a character recognition rate on an obliquely picked-up image by converting the obliquely picked-up image into a normal vision image without physically rotation-operating a recognition object. CONSTITUTION:A correction device for picked-up image is constituted by a frame memory 7 storing the picked-up still image, an address register 3 for accessing one arbitrary image element on the frame memory 7, an address computing element 2 and an image data bus 5. The address of a normally used vision frame memory is sequentially generated in an address counter 1 and the address is converted into the address on the obliquely used frame memory 7 in correspondence with an operation expression based on the photographing condition of a distance between an image pickup device and a subject and an angle in the address conversion device 2. Thus, data is read from the memory 7 by the converted address.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a captured image correction device, and more particularly, to an image correction device that optically captures an image of a medium to be recognized in character recognition from diagonally in front of the medium, and once stores the captured image in an image memory.
The present invention relates to an oblique image correction device used in a character recognition system that recognizes characters on a medium based on the image information.

0002

[Prior Art] The imaging section of a conventional optical character recognition device is one in which the distance between the scanner and the recognition medium is sufficiently small compared to the size of the device, such as in an automatic postal code reading device, or the distance between the scanner and the recognition medium is sufficiently small compared to the size of the device. Since the object was a postcard, which could be easily rotated manually inside the device, oblique imaging distortion caused by imaging from an oblique angle did not occur. Therefore, no correction operation was necessary.

[0003] Generally, the character recording medium that is the object of character recognition has a planar shape, and in order to recognize characters only from optical image information, it is necessary to image the recording medium from the front.
This is the most advantageous in terms of recognition performance.

However, vehicle license plates, etc.
When it is impossible to manually rotate the character recording medium inside the device, such as when recognizing a large object moving at high speed, or when the distance between the imaging device and the recognition medium is too short compared to the size of the device. In cases where it is impossible to rotate the recognition target, such as when the object is large, it is preferable to mathematically transform the coordinates of the actual captured image to obtain a pseudo-rotation-corrected image.

[0005] This is because character dictionaries installed in normal recognition devices are intended for images viewed from the front, and not for images taken from an oblique angle. If an attempt was made to create a dictionary for oblique images, it would be necessary to create a dictionary according to each individual imaging condition, which would result in a huge dictionary.

[0006]

OBJECTS OF THE INVENTION The present invention has been made to solve the above-mentioned conventional drawbacks, and its purpose is to provide a captured image correction device that can improve character recognition performance.

[0007]

[Structure of the Invention] A captured image correction device according to the present invention includes a frame memory for oblique view that holds an oblique image taken from an oblique direction, and a frame memory for normal view that holds a front view image when viewed from the front. A captured image correction device comprising: address generation means for sequentially generating addresses of the frame memory for normal vision; The present invention is characterized by comprising: address calculation means for converting into an address on the diagonal frame memory according to an arithmetic expression based on the calculation formula; and reading means for reading data from the diagonal frame memory using the converted address. do.

[0008]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Next, the present invention will be explained with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of the main parts of an embodiment of a captured image correction apparatus according to the present invention. In the figure, the oblique captured image correction device according to an embodiment of the present invention includes a frame memory 7 for storing captured still images, an address register 3 for accessing any one pixel on the frame memory 7, and an address register 3 for accessing an arbitrary pixel on the frame memory 7. It includes an address calculation device 2 for sending an appropriate value to a register 3, and an image data bus 5 for transferring image information on a selected frame memory 7 to a recognition device. In the figure, 6 is a control bus, 4 is an address bus,
8 is a bus controller.

Referring to FIG. 2, in order to convert the oblique image D (x', y') into a normal view image N (x, y), each of the constituent elements of the oblique image D is It is sufficient to place the pixels at corresponding positions on the normal vision image N. However, in the digital image, when each pixel of the oblique image D is embedded into the normal image N, each pixel in the oblique image D and the normal image N is not in a 1:1 ratio. Generally, a blank space is created in the normal view image N.

[0011] Therefore, each pixel on the en face image N is raster scanned in order, and the oblique image D corresponding to that pixel is obtained.
By filling the density level of the upper pixel into the normal view image N, it is possible to obtain a pseudo normal view image and to eliminate empty spaces in the image.

Returning to FIG. 1, in order to perform such raster scanning, the apparatus of this embodiment is provided with an address counter 1. In other words, in order to implement raster scanning using an electrical circuit, the address counter 1 increments the pixel count values x, y in the horizontal and vertical directions so that the point (x, y) raster scans the screen. . At this time, the range of change in x and y is defined as width W and height H.
If the image is limited by , it is possible to obtain only the necessary partial image of the en face image.

In such a configuration, address counter 1
The x and y values generated in step 2 are output to the address arithmetic unit 2, and the calculations of equations (1) and (2) are executed. In other words, by calculating these equations (1) and (2), the point (x,
y) to a point (x', y') on the obliquely captured image.

[0014] x'=f(x, y, α, θ)...(1) y'=g(x,
y, α, θ)...(2) In equations (1) and (2), θ
is the angle between the imaging device and the subject in the horizontal direction, and α is the angle in the vertical direction, which is uniquely determined by the imaging conditions. Also, (x', y') is (x, y
), and represents a specific pixel.

Note that in equations (1) and (2), the functions f,
g has various imaging conditions as parameters. An example of the function f is shown in equation (3), and an example of the function g is shown in equation (4).

x'={(-d sin α sin γ-d
cosθ cosα cosγ)x+d sinθ
cos α sin γ}/{(- sin γ)y+d s
inα sinγ+dcosθ cosα cosγ}
...(3) y'={(-d sinα sinγ-
d cosθ cosα cosγ) (cosγ)y
+(sin γ)y(d sinα cosγ+d
cosθ cosα sinγ)}/{(- sinγ
)y+dsinα sinγ+d cosθ cosα
cos γ} (4) In equations (3) and (4), γ is the angle between the recognition target plane and the vertical axis, and is stored as a constant in the device. Further, d is the distance from the imaging device to the subject.

In other words, by calculating a function based on the distance between the imaging device and the subject, the imaging angle, etc., the address of the normal view image is converted into the address of the perspective view image.

The address arithmetic unit 2 receives θ, α, d, and γ as input, executes calculations according to the above equations (1) and (2), and outputs the calculation results to the address register 3.

Address register 3 is a buffer for electrically holding the address value on address bus 4 for a certain period of time.

On the other hand, the frame memory 7 temporarily stores the obliquely captured multivalued image, but at the same time constantly monitors the address bus 4 and stores the addresses (x',
y') is sent to the image data bus 5. In other words, data is read from the address specified by the address register 3. At this time, the control bus 6 plays the role of transmitting the enable states of the address bus 4 and image data bus 5 from the bus controller 8 to the frame memory 7 and the address register 3.

When the address register 3 and frame memory 7 output data onto the bus, they apply for output permission to the bus controller 8 via the control bus 6. At this time, the bus controller 8 adjusts the data output timing between the two.

As a result, corrected image information of width and height (W, H) is obtained on the image data bus 5 in accordance with the imaging conditions. This corrected image information is held in a frame memory for front view images (not shown). Then, by sending this corrected signal to a recognition device (not shown), it is possible to perform dictionary recognition of the stereoscopic image. Therefore, character recognition can be performed by simply preparing a dictionary of front-view images. In addition, the signal is branched through a distributor,
By superimposing a synchronization signal, a front-view image can be displayed on a monitor.

[0023]

Effects of the Invention As explained above, the present invention is capable of converting an obliquely captured image into a straight-view image without physically rotating the recognition target. Even in devices that try to recognize media,
This has the advantage that it is possible to create a device with recognition performance comparable to that of images captured with normal vision.

[Brief explanation of the drawing]

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

FIG. 2 is a conceptual diagram showing the correspondence between an oblique captured image and a normal view image.

[Explanation of symbols]

1 Address counter 2 Address calculation device 3 Address register 7 Frame memory 8 Bus controller

Claims (1)

[Claims] 1. A captured image correction device comprising an oblique frame memory for holding an oblique image taken from an oblique direction, and a normal view frame memory for holding a normal view image when viewed from the front. address generation means for sequentially generating addresses of the frame memory for normal vision; and address generation means for sequentially generating addresses of the frame memory for normal vision; 1. A captured image correction device comprising: address calculation means for converting the address into an address on the diagonal frame memory; and readout means for reading data from the diagonal frame memory using the converted address.