JPH06251128A - Image processor and copying machine using its processor - Google Patents

Abstract

PURPOSE:To provide an image processor which can detect at a high speed paper money, etc., being an object by a single scan. CONSTITUTION:Prescribed image information is inputted successively to a specific pattern detecting means 1, and by its detecting means, gradated image data is formed by lowering the resolution by an averaging processing part 10, and by an edge extraction filter 13 of the next stage, an edge of density/ color tone is extracted from image data of low resolution. Subsequently, the extracted edge is sent to an angle candidate rough retrieving/positioning part 15, expanded to a built-in register buffer array, and whether two pieces of orthogonal straight lines (corners) exist or not is detected by a logic circuit connected thereto. When they are detected, data in a prescribed area containing them is given to a segmenting means 2, a necessary part is segmented therein and outputted to a specific pattern matching means 3 of the next stage, and whether a candidate pattern segmented therein is a specific pattern or not is decided.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus suitable for preventing forgery of banknotes, securities, etc. and a copying machine equipped with the image processing apparatus.

[0002]

2. Description of the Related Art With the recent development of a copying machine such as a full-color copying machine, the quality of a copied image has reached a level indistinguishable from the original image and the naked eye can easily obtain such a faithful copy. . However, with it, banknotes,
It is necessary to consider that the risk of being illegally used for counterfeiting is increased even though copying of securities or the like is originally prohibited socially, and various anti-counterfeiting devices have been developed to prevent such risk. Then, as one of them, for example, there is an image processing apparatus disclosed in Japanese Patent Laid-Open No. 210481/1990.

That is, such a processing apparatus is mounted on a full-color digital copying machine that scans the entire surface of an original four times to read the original image placed on the original table and performs copying processing. When a banknote or the like is placed on the table, the rough position where the banknote is likely to exist is detected based on the watermark of the banknote to be detected for the forgery prevention by the first scan, and the second time is detected. At the time of scanning, the accurate position of the banknote and the angle at which the banknote is placed (correct position coordinate of the banknote on the document table) are detected. Then, at the time of the third scan, the position coordinates of the red stamp printed on the banknote are calculated from the accurate position of the banknote obtained at the second scan, and the image of the area in which the red stamp exists is calculated based on the calculated position coordinates. It is configured to detect that the object to be detected is placed on the platen by performing a plurality of scans such as extracting and determining whether or not it is a red stamp. And
If it is detected that a bill or the like is placed on the platen and is about to be copied, a predetermined forgery prevention such as displaying the entire screen in black or prohibiting copying during the fourth scanning is performed. It is supposed to process.

[0004]

However, in the above-mentioned conventional apparatus, it is necessary to scan a plurality of times in order to detect an object such as a banknote for which copying is prohibited, and it takes a long time for the determination. In addition, a memory capacity corresponding to the size of a bill is required, which not only increases the cost of the device,
There are few types of bills that can be discriminated (detected). Further, in color copying machines and the like, in addition to the above four-time scanning method, there are also three-time and one-time scanning methods.
The above processing apparatus cannot be applied to the copying machine of this type. Moreover, a very large memory capacity corresponding to at least the size of the bill or the like to be detected is required. This inevitably causes a problem that the types of bills and the like that can be detected are reduced.

The present invention has been made in view of the above background, and an object of the present invention is to detect a bill or the like as an object with one scan and to use a small memory capacity. Another object of the present invention is to provide an image processing apparatus that can be processed at high speed in real time and can be manufactured at low cost, and a copying machine using the same.

[0006]

In order to achieve the above object, in an image processing apparatus according to the present invention, a means for detecting a specific pattern-like pattern existing in given image information, and the detected pattern. From the image information, and means for calculating the similarity between the preset reference pattern and the cut-out pattern, and the detecting means at least extracts an edge from the image information. The specific pattern-like pattern is detected based on the edge extracted by the extracting means.

Further, the detecting means is connected to at least means for calculating binary information from image information, means for holding the calculated binary information of the plane in a register buffer, and each output of the register buffer. It may be configured by a predetermined logic circuit for detecting the specific pattern-like pattern.

[0008] Preferably, the means for detecting the pattern further comprises means for forming a blurred image by reducing the resolution of the given image information, and performing the above processes on the blurred image. To do it.

Further, in the copying machine according to the present invention, the above-mentioned various image processing apparatuses are converted into signals for printing at least a document reading means and a reading means connected to the reading means. It is mounted on a copying machine equipped with a color signal conversion unit and a printing unit that receives an output from the color signal conversion unit and performs a predetermined printing process. Then, the image data output from the document reading unit is input to the image processing apparatus in parallel with the color signal conversion unit, and the image processing apparatus registers in advance that the object being copied is a banknote or the like. Judge whether it is a predetermined object,
At least when it is determined that the object is the predetermined object, a control signal is sent to a predetermined processing means of the copying machine to suppress copying.

[0010]

The edge is extracted from the image information given by the detecting means based on the characteristic amount such as the density, and the plane (two-dimensional) information is converted into the line (one-dimensional) information. Then, it is detected whether or not there is a predetermined pattern in the image information given based on the line information. Then, in some cases, a cutting-out unit at the next stage cuts out a predetermined area (area including the detected pattern) in the image information,
It is sent to a means for calculating the degree of similarity, and predetermined arithmetic processing is performed there to obtain the reference pattern likelihood (similarity). If the calculated similarity is high, the detected pattern is recognized as the specific pattern which is the final detection purpose. Since various processes are performed based on the line information, the above processes are performed at high speed. In addition, when the detection process of a specific pattern is performed by forming a register buffer and a predetermined logic circuit, the process is performed at higher speed. Processed and processed in real time for reading image information.

On the other hand, when such an image processing apparatus is installed in a copying machine, a specific pattern (reference pattern) is set so as to correspond to an image of a predetermined area of a copy-prohibited object such as a banknote, so that copying is performed. When a copy-prohibited object is to be copied using a copying machine, the image processing apparatus is operated in parallel with the normal copy processing so that the copy (original) to be copied contains a predetermined specific pattern. If it is determined that there is a predetermined copy prohibition process is executed. Then, as described above, the image processing apparatus is processed at high speed and follows the reading of the document in real time.

[0012]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT A preferred embodiment of an image processing apparatus according to the present invention and a copying machine using the same will be described in detail with reference to the accompanying drawings. In this example, when an image is mounted on a full-color copying machine and an attempt is made to copy a banknote or the like for which copying is prohibited using such a full-color copying machine, image processing for detecting that and stopping the copying process. The device is shown. That is, as shown in FIG. 1, image information read by an image sensor such as a CCD, which is an optical reading device provided in the main body of the copying machine, is input to the specific pattern detecting means 1. This image information is sequentially sent in real time by a predetermined area as scanning by the image sensor progresses. Specific data is full-color information of red (R) and green ( 400) for each of G) and blue (B) components
The resolution is DPI.

With respect to the blurred image formed by compressing the input image data by the specific pattern detection means 1, whether there is a candidate pattern that seems to be the specific pattern to be detected in the area by pattern matching. While checking whether or not the reference position of such a candidate (in the case of a rectangular shape, the position of the apex of a corner or the like) is specified, the data is sent to the specific pattern cutting means 2 in the next stage.

Then, the specific pattern cutting means 2
Then, with respect to the blurred image data including the candidate pattern detected by the rough search, a predetermined portion around the blurred image data is cut out based on the reference position, and a comparison pattern for performing pattern matching in a subsequent process is created ( Actually, the feature amount is extracted) and is sent to the specific pattern matching means 3.

Then, the specific pattern matching means 3 obtains the matching degree of the comparison pattern to the reference pattern (specific pattern) by fuzzy pattern matching,
It is determined whether or not the comparison pattern is a specific pattern, and a predetermined control signal is output at least when the comparison pattern is the specific pattern.

That is, in this example, the read data of a normal copying machine is subjected to pattern matching on the basis of a pattern created by blurring the reference pattern to successively perform a relatively rough rough search, which seems to be a specific pattern. A candidate pattern is extracted, and when such a candidate pattern is detected, pattern matching by fuzzy inference is sequentially performed on the candidate pattern to determine whether or not it is a specific pattern. Moreover, since the detection and cutting of the pattern are processed at high speed by hardware as described later, the processing can be performed in real time without using a storage unit having a particularly large storage capacity.

Next, a concrete configuration of each of the above means will be described. First, the specific pattern detection unit 1 includes a plurality of small pixels of 400 DPI sent from an image sensor (not shown) in the averaging processing unit 10 (N pixels in the main scanning direction and M lines in the sub scanning direction). ) Are summarized as one, and their densities and / or color tones (hereinafter, basically explained about densities) are averaged to create slightly large (low resolution) grayscale image data. Here, since the image data given is for each line in the main scanning direction,
In the averaging processing unit 10, the result obtained by adding the densities of the predetermined N pixels in each line using the built-in adder is connected to the averaging processing unit 10. Processing such as storing the data in the line memory 11 or calling already stored data on the previous line, adding the density data of the line currently being processed to it, and rewriting the data is performed. Then, when the total sum of the image data for a predetermined number of M lines has been obtained, division is performed by a built-in divider to obtain an average value, and the average value is stored in the second line memory 12.

As a result, the image stored in the second line memory 12 becomes a blurred image as compared with the original 400 DPI image data, and a fine pattern disappears and a rough shape appears. Therefore, even if the original image read and input has a print deviation or the like and has a slight difference from the specific pattern stored in advance, the difference caused by blurring the image is eliminated (erased). . As a result, various processes such as various pattern detection and pattern recognition thereafter can be performed at high speed and accurately.

Then, the low-resolution grayscale image data stored in the second line memory 12 is sent to the edge extraction filter 13 in the next stage, where a filter having a predetermined filter coefficient of 3 × 3 is formed. Is used to extract the density edge and / or color tone edge of the grayscale image. As the filter coefficient used here, for example, the target pixel is "+8" and the surrounding eight pixels are "-1", or the target pixel is "+ 8 + N" and the surrounding is "-1". It is set as appropriate depending on the object to be detected and the like. As the data to be input to the edge extraction filter 13, R, G and B may be directly input, or R-G and G-.
Various types such as B, B-R, R + G + B, and a linear format in which RGB are appropriately weighted can be used, and these are also determined according to the detection target. Then, the predetermined edge-extracted image data is written in the third line memory 14.

Further, the image data stored in the third line memory 14 is successively sent to the corner candidate rough search / positioning unit 15 in the next stage, and whether or not there is a candidate pattern which seems to be a specific pattern to be detected there. If such a candidate is detected, the position and angle of the apex of the corner, which is the reference position, are specified, and such data is sent to the specific pattern cutting means 2 in the next stage. The image data stored in the third line memory 14 is also supplied to the specific pattern cutting means 2.

The corner candidate rough search / positioning unit 15 will be described. A diagram set from the third line memory 14 using a straight line extraction mask (black or shaded portion in the drawing) as shown in FIG. 31 × 31 as shown in 3
It is adapted to detect whether or not there are two straight lines intersecting at a predetermined angle (90 degrees) at the center position in the data in the shift register (register buffer).

First, the straight line extraction mask will be described.
In FIG. 2, for convenience, 90 ° of the first quadrant is shown, and in addition, in order to clearly show each adjacent straight line, it is divided into four parts of FIGS. Therefore,
The actual mask in the first quadrant is a combination of FIGS. 2A to 2D, and includes 30 masks 0 to 29. Then, similar masks are prepared for the remaining 270 degrees, and a total of 120 masks are provided, resulting in a resolution of 3.8 degrees.

Basically, when the extracted edge matches any one of the straight line masks described above (the pixels forming the edge are located on the pixels forming the straight line mask), the edge is a straight line. It is supposed to be. However, in this example, in consideration of detection error and the like, when there is no blank (non-edge portion) in two or more adjacent pixels that are adjacent to each other among the pixels that form each linear mask,
It is determined that there is a straight edge on the straight mask.

Further, in the case where there are a plurality of straight lines thus detected, the center position of the 31 × 31 shift register has a specific pattern (candidate pattern) when the angle formed by two arbitrary straight lines is 90 degrees. It is the coordinates of the corner apex. Further, the angle (orientation) at which the corner is inclined can be detected by the value of the detected line mask number n.

Next, an actual circuit configuration for performing the above various detections will be described. As shown in FIG. 4, of the pixels (actually, the outputs of the registers) forming the masks of the straight lines, the adjacent two pixels are connected to the respective input terminals of the 2-input straight line detecting OR element 16. The outputs of all the linear detection OR elements 16 are input to the multi-input linear detection AND element 17. Accordingly, if there is an edge on a pixel forming a predetermined mask, the output of the corresponding straight-line detecting AND element 17 becomes "1".
A straight line can be detected at high speed.

In this example, the data for four pixels from the center is not used for detecting a straight line, but this is because a plurality of straight lines pass over one pixel in the region of four pixels from the center. It is not very important as a feature quantity for identification,
For example, there is no near vertices (not connected at vertices)
This is because it is possible to detect even two straight lines (the angle formed is 90 degrees). In addition, the rule for extracting a straight line is not limited to the one described above (in this example, a straight line is used if a plurality of consecutive pixels are not white), and the detection target (specific pattern)
It may be changed appropriately according to the above, and in such a case, it may be dealt with by reassembling the logic circuit accordingly.

As a means for detecting whether two straight lines are orthogonal to each other, as shown in FIG. 5, the output of the n-th straight line detecting AND element 17 is a 2-input right angle determining AND element 18.
One of the input terminals is connected to one input terminal, and the other input terminal of the AND element 18 for determining the right angle is connected to n + 29th, n + 30th, and n + th.
AND element 17 for detecting the straight line for each of the 31st straight line masks
Is input through the right angle determination OR element 19. Then, the right angle determination AND element 18
And a right angle determining OR element 19 are provided in a predetermined number, and are connected to the outputs of the elements 17 corresponding to all combinations of the straight line detecting AND elements 17 which should be right angles.

That is, in this example, since 30 straight line masks are provided for 90 degrees, two straight lines corresponding to the nth and n + 30ths form an angle of 90 degrees, so 31 × 31.
If there are two straight lines orthogonal to each other in the center pixel in the area of, the output of the corresponding predetermined orthogonal determination AND element 18 becomes "1". Therefore, the detection of two orthogonal straight lines existing in the image data in the 31 × 31 area is performed by a gate circuit that is appropriately combined and connected,
Moreover, since the detection processing is executed by the circuit composed of hardware, it is possible to make an instant determination.

Furthermore, the inclination angle of the corner can be automatically determined by specifying the orthogonal determination AND element 18 whose output is "1". Note that it is not only n + 30 that is input to the AND element 18 in correspondence with nth, but n
+29 and n + 31 are also entered
This is because it corresponds to the first detected angle error.

Next, the third line memory 1 will be described with reference to FIG.
A device for writing data from 4 to 31 × 31 shift registers will be described. As shown in the figure, first, an 8-bit shift register 2 with parallel input and serial output is provided on the input side.
31 columns of 0s are arranged in a predetermined positional relationship. Specifically, if four rows are arranged with the front and back aligned,
4 columns are arranged in the state of moving forward by 1 bit in the shift direction, and then repeated. And each shift register 2
The serial input / parallel output shift register 2 having a predetermined bit length of 38 to 31 bits depending on the shift of the input side shift register 20 in the shift direction.
1 is connected. As a result, each shift register 21
And the final stage of the match, and 31 × 31 from the final stage side
Area is formed, and each of the linear detection OR elements 16 is connected to each register, which is a parallel output in the area, through a predetermined wiring.

Then, the shift register 20 having such a configuration,
Data writing to 21 is performed using the 8-bit DF / F 22 connected to the third line memory 14 and the load flag generation circuit 23. That is, DF / F
The 8-bit output of 22 is all 8 of the 31 columns on the input side.
It is connected in parallel to the bit shift register 20 (in the figure, for convenience, only the uppermost one column is connected (indicated by an arrow)), and the load flag generation circuit 23 outputs a flag (output is "1") shift register 20
The output of the D-F / F22 is written in. Then, in this example, the load flag generating circuit 23
Also uses a cyclic parallel output shift register having a predetermined number of bits, and data in which only one position is "1" and the rest is "0" is circulated.

Furthermore, the shift register constituting the load flag generating circuit 23 generates four pulses during one pixel scan (while one pixel of low resolution image data is written in the third line memory). The shift registers 20 and 21 are shifted by a shift timing signal (STS) 2 which generates a pulse once during one pixel scanning. .

As a result, the load flags are sequentially set from the load flag generating circuit 23 to the shift registers on the input side of the first to fourth columns from the top while one pixel is scanned, so that the third line memory 14 The data for the predetermined past 8 pixels stored in 8 is written in the 8 × 4 area A. If the above 1 pixel is scanned, S
TS2 generates a pulse once, and the data written in the area A is shifted by one pixel. Then, while the next one pixel is scanned, the load flags are sequentially set from the load flag generation circuit 23 to the shift registers on the input side of the 5th to 8th columns from the top, so that they are stored in the third line memory 14. 8 × related to the data of the predetermined past 8 pixels
4 is written in area B.

At this time, as described above, the area A and the area B are displaced by one pixel, and the area A is
Since the data written in the area has already been shifted by one pixel during the writing process in area B,
After all, with the pixel written in the beginning B1 of this area B,
The pixels written in the area A and shifted by one and written in the head A1 are at the same position in the main scanning direction (shifted by a predetermined number of columns in the sub scanning direction) in the original image.

By performing the writing process in the following order, in the 31 × 31 area formed in the shift register 21 in the subsequent stage, the averaged low image data in the relative positional relationship is set. Predetermined linear detection OR elements 16 in which predetermined regions of
Is input to and the orthogonal straight line is detected.

On the other hand, the specific pattern cutting means 2 is based on the position of the apex of the corner and the inclination angle θ (calculated from the detected linear mask number n) detected by the corner candidate rough search / positioning unit 15. Then, the effective pixel area in the image data of the corresponding predetermined area (area including the detected corner) given from the third line memory 14 is extracted.

That is, in many cases, the search target area R (a square area in which eight feature amount blocks composed of 8 pixels × 8 pixels are provided on one side) as shown in FIG. 6 is read by the image sensor. The rectangular area T is inclined by a predetermined angle θ with respect to the Cartesian coordinate systems x and y with reference to the main scanning / sub scanning directions. Therefore, the above-mentioned orthogonal coordinate system is subjected to coordinate conversion to an orthogonal coordinate system x ', y'consisting of two adjacent sides of the search target area, and at the same time, it is determined whether or not the search target area R is an effective pixel. Only the extracted feature amount is extracted, and the feature amount for the next recognition process is extracted. The specific processing procedure is as shown in the flow in FIG.

That is, first, from the vertex position coordinates (X0, Y0) of the given corner and the inclination angle θ, the diagonal of the circumscribed square (indicated by the broken line in the figure) of the search target area as shown in FIG. 6B. The coordinates of the starting point and the ending point in the Cartesian coordinate system of x and y which are the two vertices of are obtained (S101). And
As a concrete calculation method, first, the quadrant is obtained from the angle θ (in this example, θ has a resolution of 128, so
The bit determines the quadrant of θ, and the angle (address) in the quadrant is determined by the lower 5 bits) and the range of the angle θ (quadrant), and is obtained from the following table.

[0039]

[Table 1] 0 <θ <π / 2 Base point: (X0-64sin θ, Y0) End point: (X0 + 64cos θ, Y0 + 64cos θ + 64si
n θ) π / 2 <θ <π Reference point: (X0 + 64cos θ + 64sin θ, Y0 + 64co
s θ) End point: (X0, Y0 + 64sin θ) π <θ <3π / 2 Base point: (X0 + 64cos θ, Y0 + 64cos θ + 64si
n θ) End point: (X0-64sin θ, Y0) 3π / 2 <θ <2π Base point: (X0, Y0 + 64sin θ) End point: (X0 + 64cos θ + 64sin θ, Y0 + 64co
s θ) If the base point and the end point are calculated in this way, the bytes of the memory containing the base point and the end point are set as the base point byte and the end point byte, respectively, and the target byte is set to the base point and the target bit is set to 0 by default. Yes (S102). Then, the converted pixel position is calculated from the actual image position (S10).
3). That is, the coordinates (x, y) in the coordinate system of the actual image,
The relationship of the coordinates (x ', y') in the coordinate system after conversion is x '= xcos θ + ysin θ y ′ = − xsin θ + ycos θ, and thus can be easily obtained by performing such coordinate conversion.

Next, a feature amount block is calculated from the post-conversion pixel position, and it is further determined whether or not the pixel currently being processed is a feature amount block (further, which feature block), and an effective pixel in the search target area. (See FIG. 6B)
Feature calculation, that is, if the pixel being processed is black,
After performing the process of adding 1 to the corresponding feature amount block,
The target bit is incremented and the invalid pixel (see FIG. 6B)
Reference), the target bit is incremented as it is (S104 to 107). Here, in the case of this example, whether the pixel is a feature amount block or not and which feature amount block the pixel belongs to is determined as the search target region R is a square having 64 pixels on one side. When the coordinate values x ', y'in 0 are 0 to 63, it can be determined that the pixel is an effective pixel and the other pixels are ineffective pixels. Further, in the case of an effective pixel, which feature amount block is included in the coordinate value is obtained. That is, for example, x'of the coordinates (x ', y'),
If y'is both in the range of 0 to 7, it can be easily obtained from the coordinate values, such as the first feature amount block.

Then, the above steps 103 to 107 are repeated until the number of bits to be processed reaches 8. When the number of bits reaches 8, it means that the pixel becomes the next byte in the rectangular area T. The target bit is initialized to 0 and the target byte is incremented. Then, the target byte is the end of the line (base point +12 in the example of FIG. 6)
If not, the process returns to step 103 and 0 of the target byte
The feature amount is calculated based on the pixels for 7 bits. If the target byte is at the end of the line, the process moves to the first byte of the next line, and the process of steps 103 to 111 is performed for that line in the same manner as described above to calculate the feature amount (S1).
08-112). As a result, the feature amount extraction for all effective pixels is completed, and the extracted feature amount is written in the fuzzy pattern memory 30 in the next stage. In addition,
The feature amount in this example is the number of black pixels existing in each feature amount block (8 × 8 pixels).

As a specific configuration for performing the above processing, the hardware shown in FIGS. 8 and 9 is used. First, each ROM stores the values of cos θ and sin θ corresponding to 90 degrees corresponding to 128 resolutions.
That is, the relationship between the coordinates (x ′, y ′) after conversion and the coordinates (x, y) before conversion is explained by dividing it into the value of θ (quadrant).

0 <θ <π / 2 x ′ = xcos θ ′ + ysin θ ′ y ′ = − xsin θ ′ + ycos θ ′ θ ′ = θ π / 2 <θ <π x ′ = − xcos θ ′ + ysin θ ′ y ′ = − xsin θ′−ycos θ ′ θ ′ = θ−π / 2 π <θ <3π / 2 x ′ = − xcos θ′−ysin θ ′ y ′ = xsin θ′−ycos θ ′ θ ′ = θ−π 3π / 2 <θ <2π x ′ = xcos θ′−ysin θ ′ y ′ = xsin θ ′ + ycos θ ′ θ ′ = θ−3π / 2

[0044]

Cos θ = −sin (θ−π / 2) = − cos (θ−π) = sin (θ−3π / 2) sin θ = cos (θ−π / 2) = − sin (θ− Since π) = − cos (θ−3π / 2), if the values of cos θ and sin θ from 0 to π / 2 are prepared, 0 can be obtained by adding a negative number to the values as necessary. This is because x ′ and y ′ up to 2π can be obtained.

Therefore, as shown in FIG.
To the coordinate value of the target pixel before conversion and the lower 5 bits of θ (corresponding to 0 to 90 degrees (the upper 2 bits represent the quadrant)), the corresponding xcos θ, xsin θ, ycos θ, ysin θ are output. Therefore, the coordinates after conversion can be obtained by appropriately adding and subtracting the respective output values, but in this example, in order to further obtain the coordinate values after conversion of pixels adjacent in the x direction, D-FF and With the circuit configured by Adder1, (x + n) co based on the pixel next to the target pixel
n (n = 0 to 7) in s θ, (x + n) sin θ is calculated (more strictly, ncos θ and n sin θ are calculated),
The above output xcos θ, using Adder2,
Calculated by adding to xsin θ. Then, the result of the calculation is used for each multiplexer (4-1M) having four inputs and one output.
PX). In addition, −cos θ, −sin
In order to obtain θ, the output of Adder2 above
It is adapted to be input to the 4-1MPX via t.

Further, the output of the 4-1MPX is selected by the upper 2 bits of θ given thereto,
By adding the output of -1MPX using Adder3, the converted coordinates (x ', y') can be obtained. That is, as described above, since the upper 2 bits of θ indicate the quadrant, they are selected and output so as to be a conversion formula corresponding to the above quadrant.

The converted coordinate data (x ', y') thus obtained are input to the circuit shown in FIG. That is, FIG. 9 configures a feature amount calculation circuit, and in this example, there are 64 feature amount blocks. Therefore, 64 such circuits are also provided, and the output of the circuit of FIG. It is designed to be input.

Then, the input transformed coordinates (x ',
y ') is by Address Matching,
Whether or not there is a corresponding coordinate is determined, and if there is (that is, the coordinate within the feature amount block), the counter is incremented by 1 when it is an edge. As a result, if it is an effective pixel, it matches the Address Matching of any one of the feature amount calculation circuits, and if it is an invalid pixel, the Address of all feature amount calculation circuits.
It will not match Matching. Then, the count value of each counter becomes the feature amount of the corresponding feature amount block and is output to the matching unit 3 in the next stage as described above. As mentioned above,
Since the specific pattern cutout unit 2 is composed of hardware such as a gate circuit, extremely high-speed processing is possible, and coordinate conversion and feature amount calculation are performed in real time.

Next, the specific pattern matching means 3 will be described. The feature amount data in the effective pixel area for one extracted candidate pattern stored in the fuzzy pattern memory 30 from the specific pattern cutout unit 2 is sent to the normalization unit 31, where density correction,
The input data for performing the fuzzy inference of the next stage is created by performing the average value correction and various other processes. Then, the input value data of the inference thus created is sent to the fuzzy inference unit (FP-5000: fuzzy chip manufactured by OMRON Corporation) 33 in the next stage, and the memory 34 is stored therein.
Inference processing is performed based on fuzzy knowledge such as rules and membership functions stored in, and the degree of similarity of given image data with a preset specific pattern is determined. Then, the inference result is output to the main body of the copying machine. When the degree of conformity exceeds a certain threshold, the copier body determines that the copy is prohibited such as banknotes and performs a predetermined process for preventing forgery (copy prohibition, output to the entire black screen, etc.). ing. The image processing apparatus may also determine whether or not the bill is a prohibited copy object (specific pattern) and output the determination result (stop signal or the like). In addition, the recognition (similarity finding) process does not necessarily need to use the above fuzzy pattern matching, and various methods can be adopted.

In the above example, the specific pattern is
Since the formed angle includes a 90-degree corner, the angle detected by the register buffer and the predetermined logic circuit is also a straight line, and the formed angle is 90 degrees. However, the present invention is not limited to this. May be a predetermined angle other than 90 degrees, and the detection target is not limited to a straight line, and may be a circle, a curved line, or any other shape. However, in that case, the mask is set according to the shape.

FIGS. 10 and 11 show an example in which the above apparatus is actually incorporated in a copying machine. As shown in the figure, the reflected light of the light emitted from the lamp 52 on the original 51 placed on the original table 50 is read through the optical system 53 by the CCD 54 which is the image sensor to read the image of the original. It should be noted that the lamp 52 and the plane mirror or the like forming the optical system 53 move at a predetermined speed to perform scanning to scan a predetermined portion of the document 51 at a C position.
The CD 54 is sequentially read and image data (R, G, B) is sent to the signal processing unit 55.

As shown in FIG. 10, the signal processing unit 55 is equipped with a normal color processing circuit 56 and the above-described image processing device 57 according to the present invention, and the above image data includes the color processing circuit 56 and the image. The data is sent to the processing device 57 in parallel. Then, in the color processing circuit 56, the components are decomposed into magenta (M), cyan (C), yellow (Y) and black (Bk) components and output to the printing unit 58.
Then, actually, scanning is performed four times, and one component out of the four components (M, C, Y, Bk) is output to the laser driver 59 arranged on the input side of the printing unit 58 with one scanning. Then, the laser light is applied to a predetermined position of the photosensitive drum 60. Then, after the four scans are completed, the copy paper is copied, and a copy 61 is obtained.
Is output. Since a mechanism for performing a specific copying process is the same as that of the conventional one, its description is omitted.

On the other hand, in the image processing device 57, while the signal processing in the color processing circuit 56 is being performed, the above-described processing is performed in parallel with the signal processing, and the specific (reference) pattern of the pattern in the image data being read If the degree of similarity is determined and the original document 51 being read is a copy-prohibited object such as a banknote, a control signal for stopping the output of the laser driver 59 is issued, or a control signal is sent to the color processing circuit 56. Is sent to perform various copy prohibition processing such as making the entire copy screen a black image.

In this example, since the determination processing can be performed in real time by one scan, it is possible to support the mounted copying machine of the one-scan type. Then, not only the entire bill but a part of the specific pattern is detected and pattern matching is performed, so that various storage capacities are small, and even if it corresponds to various kinds of forgery prevention as described above, the whole As a result, a large memory capacity is unnecessary.

Although the above-described embodiments have been described as being applied to a copying machine, the present invention is not limited to this, and is applicable to, for example, a color printer, a FAX, a communication transmission device and various other devices. Of course.

[0056]

As described above, in the image processing apparatus and the copying machine using the same according to the present invention, one of the objects of the copy (printing) prohibited object is irrespective of the size of the whole (outer shape). Since a specific pattern included in a part is focused and detection is performed depending on whether or not the specific pattern is included, the memory capacity used is small and the cost is low. Moreover, since the detection of the specific pattern to the determination of the similarity can be performed in real time by one scan, a high recognition rate can be obtained without oversight even in a one-scan type copying machine. Then, since various types of processing are recognized by performing various types of processing based on the planar line information obtained by edge extraction, the image data and range to be checked can be reduced compared to the case of surface information. The speed can be increased, and the device and the memory capacity used can be downsized.

Further, when a plane image is developed on the register buffer and the detection processing is performed by hardware by the logic circuit, the above effect is further enhanced by accessing the memory each time. Becomes Then, the recognition process can be followed in real time with respect to the image reading speed, and a memory like a save buffer for temporarily holding the detected candidate data becomes unnecessary, and the memory capacity can be further reduced.

In addition, when a given image is blurred at the time of detection and various processing is performed based on the blurred image (pattern), it is strong against a positional shift, dirt, a reading error, etc. In addition, it is possible to reliably and reliably extract the image data having the specific pattern.

By mounting the image processing apparatus in a copying machine, it is possible to reliably prohibit the output of the copied matter such as banknotes and securities (copying itself is not performed, original (copying It is possible to copy and output images different from prohibited items).

[Brief description of drawings]

FIG. 1 is a block configuration diagram showing an embodiment of an image processing apparatus according to the present invention.

FIG. 2 is a diagram showing an example of a linear mask.

FIG. 3 is a diagram showing an example of a register buffer in which plane image information is expanded and a unit for writing data into the register buffer.

FIG. 4 is a diagram showing an example of a circuit for detecting a straight line.

FIG. 5 is a diagram showing an example of a circuit for determining whether or not two detected straight lines are orthogonal.

FIG. 6 is a diagram for explaining the operation of a means for cutting out a pattern.

FIG. 7 is a flow chart diagram thereof.

FIG. 8 is a diagram showing a part of a configuration example of hardware for executing it.

FIG. 9 is a diagram showing the rest of a configuration example of hardware for executing it.

FIG. 10 is a diagram showing an example of a copying machine according to the present invention.

FIG. 11 is a diagram showing an example of a copying machine according to the present invention.

[Explanation of symbols]

1 Specific Pattern Detection Means 2 Specific Pattern Cutout Means 3 Specific Pattern Matching Means (Means to Calculate Similarity) 10 Averaging Processor 13 Edge Extraction Filter 15 Corner Candidate Rough Search / Positioning Unit (Register Buffer etc.)

Claims (4)

[Claims] 1. A means for detecting a pattern that seems to be a specific pattern existing in given image information, a means for cutting out the detected pattern from the image information, a preset reference pattern and the cut-out pattern. And a means for calculating the degree of similarity, and the detecting means has at least means for extracting an edge from image information, and the specific pattern-like pattern based on the edge extracted by the extracting means. An image processing apparatus, which is configured to detect a. 2. A means for detecting a pattern which seems to be a specific pattern existing in given image information, a means for cutting out the detected pattern from the image information, a preset reference pattern and the cut out pattern. And a means for calculating the binary information from the image information, and a means for holding the calculated binary information of the plane in a register buffer. An image processing apparatus comprising: a predetermined logic circuit for detecting the pattern that seems to be the specific pattern, which is connected to each output of the register buffer. 3. The means for detecting the pattern further comprises means for forming a blurred image by reducing the resolution of the given image information, and performing each of the processes on the blurred image. The image processing device according to claim 1 or 2, wherein 4. A means for reading at least a manuscript, a color signal conversion means connected to the reading means for converting the read image data into a signal for printing, and an output from the color signal conversion means, A copying machine provided with a printing means for carrying out a predetermined printing process, wherein:
The image processing apparatus according to any one of 3 above is mounted, and the image data output from the document reading unit is input to the image processing apparatus in parallel with the color signal conversion unit, and the image processing apparatus is also provided. Determines whether the object being copied is a predetermined object such as a banknote registered in advance,
A copying machine, characterized in that a control signal is sent to a predetermined processing means of the copying machine at least when it is judged that the object is the predetermined object to suppress copying.