JP3359488B2 - Image processing device - Google Patents

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 for suppressing variations in density distribution when an object being read at high speed is read by an image sensor. The present invention relates to an image processing apparatus that suppresses variation in density distribution (hereinafter referred to as “density variation”) caused by fatigue of banknotes.

[0002]

2. Description of the Related Art Conventionally, in a bill validating apparatus for recognizing bills, a bill conveyed at a high speed is read by an image sensor, and the read image data is compared with reference data prepared in advance to determine the bill being transported. Identify authenticity and denomination.

For example, Japanese Patent Application Laid-Open No. 5-143829 discloses that when a print pattern of a bill to be discriminated is read by a sensor, data in the longitudinal direction of the bill to be discriminated is calculated, and the calculated data is compared with standard data. A banknote discriminating device configured to detect continuity of a pattern in a longitudinal direction is disclosed.

[0004] Therefore, in such a bill discriminating apparatus, the accuracy of discrimination is improved by eliminating the influence of the optical system and the transport system and accurately reading the print pattern printed on the bill surface.

[0005]

However, when a bill is read at a high speed, there is a variation in density distribution due to a change in the distance between the bill conveyed at a high speed and the image sensor. It is difficult to accurately read a print pattern printed on the surface of a bill.

That is, the bill inserted into the bill insertion slot is conveyed at high speed by using a mechanical mechanism such as a conveying roller or a conveying belt. fluctuate.

[0007] Further, since such bills usually have folds, wrinkles, uneven printing, etc., the input image has more or less uneven density.

Further, there is uneven density due to variation between pixels of the image sensor.

As described above, since such density fluctuations are one of the main factors that determine the recognition accuracy in the bill validator, how to remove the density unevenness of the input image caused by the density fluctuations is described. Is an important issue.

In particular, in the case where the identification processing is performed using the characteristic amount obtained by secondarily differentiating the image data, noise and the like are further emphasized by this second differentiation, which causes an increase in the identification error. It is important to:

In view of the above, the present invention solves the above-mentioned problems, and when an object conveyed at high speed is read by an image sensor, the density distribution caused by a change in the distance between the object and the image sensor is reduced. An object of the present invention is to provide an image processing apparatus capable of suppressing variations and the like.

[0012]

In order to achieve the above object, the present invention relates to an image processing apparatus for suppressing variations in density distribution when an object during high-speed conveyance is read by the image sensor. A second differential image forming unit for forming a second differential image by secondarily differentiating the density image after smoothing the density image, and each pixel of the second differential image formed by the second differential image forming unit. Of the values
A normalization unit that normalizes the second derivative image by replacing a pixel value equal to or greater than a predetermined upper limit value with the upper limit value and replacing a pixel value equal to or less than a predetermined lower limit value with the lower limit value; Identification means for identifying the object based on the normalized second derivative image.

[0013]

Embodiments of the present invention will be described below with reference to the drawings. In this embodiment, a case is shown in which the present invention is applied to a preprocessing unit of a banknote recognition device.

FIG. 1 is a diagram showing the configuration of a bill discriminating apparatus 10 used in the present embodiment.

A bill discriminating device 10 shown in FIG. 1 is a device for reading bills conveyed at high speed by an image sensor and discriminating the authenticity and type of the bills using a Laplacian image of the read image data.

Specifically, in the bill validating apparatus 10,
Performs preprocessing to create a non-linear Laplacian image that eliminates the effects of density fluctuations and the like caused by fluctuations in the distance between the bill and the image sensor being conveyed, and performs processing based on the non-linear Laplacian image created in this preprocessing. The identification process.

That is, if the input image read by the image sensor is subjected to the second derivative, even if the smoothing is performed prior to the second derivative, the change in the distance between the banknote being conveyed and the image sensor is affected. Since the influence of density fluctuations and the like caused by the fluctuation is magnified, the bill
Creates a normalized non-linear Laplacian image by limiting each pixel value of the Laplacian image with predetermined upper and lower limits in order to remove the influence of such density fluctuations.

The reason why the bill validator 10 performs such normalization is as follows.

A Laplacian image is divided into positive and negative areas with a zero cross point as a boundary. The position of the zero crossing point where the sign changes with the density change hardly changes, and the sign of the sign does not change, but the intensity of the sign changes. Therefore, by limiting the intensity, the effect of concentration fluctuation can be minimized.

As described above, the feature of the present invention resides in the pre-processing for generating the nonlinear Laplacian image from which the influence of the density fluctuation or the like is removed. Therefore, the present embodiment is focused on the pre-processing unit 12 for performing the pre-processing. The form will be described.

As shown in FIG. 1, the bill validator 10
Are a sensor unit 11, a smoothing processing unit 12a, a secondary differentiation processing unit 12b, a normalization processing unit 12c, and a threshold storage unit 1.
It comprises a pre-processing unit 12 having 2d and an identification processing unit 13.

The second differential processing section 12b corresponds to the second differential image forming means, and the normalization processing section 12c corresponds to the normalizing means.

The sensor unit 11 is a processing unit that reads bills conveyed at a high speed by a conveying roller or the like (not shown) by an image sensor and outputs the read input image to the preprocessing unit 12.

The pre-processing unit 12 smoothes the input image to remove noise, then performs second-order differentiation of the input image to create a Laplacian image, and normalizes the Laplacian image to remove an offset component. And includes a smoothing processing unit 12a, a secondary differentiation processing unit 12b, a normalization processing unit 12c, and a threshold storage unit 12d.

Here, the smoothing processing section 12a is a processing section that blurs an input image by applying a local operator, relaxation method, processing in the frequency domain, or the like, and removes noise components existing in the input image. In the present embodiment,
A 3 × 3 local operator is used.

The secondary differential processing unit 12b is a processing unit that creates a Laplacian image by applying a Laplacian operator to the input image smoothed by the smoothing processing unit 12a. In this embodiment, 5 × 5 operators are used.

The normalization processing unit 12c generates a nonlinear Laplacian image by suppressing the Laplacian image created by the secondary differentiation processing unit 12b with the upper limit value and the lower limit value held in the threshold value storage unit 12d. It is.

That is, the normalization referred to here means that a non-linear Laplacian image in which a Laplacian image is restricted by an upper limit and a lower limit is created, and a Laplacian image whose absolute value of each pixel value is larger than necessary is equal to or smaller than a predetermined value. Means to limit.

The reason for performing such normalization is that if the above-mentioned density fluctuation exists, the second derivative value itself of the Laplacian image itself relatively fluctuates, but the positive / negative region bounded by the zero cross point of the Laplacian image is Based on little impact.

That is, the non-linear Laplacian image in which each pixel has a pixel value within a predetermined range centered on the zero-cross point by limiting the second derivative value to a certain value or more and a certain value or less is used. The effect can be eliminated.

The upper and lower limits can be arbitrarily determined based on statistics such as average and variance or various simulations.

The discrimination processing section 13 is a processing section for discriminating banknotes based on the non-linear Laplacian image created by the pre-processing section 12. More specifically, the non-linear Laplacian image is divided into blocks and a histogram is generated for each block. Calculation is performed, and the calculated histogram is compared with a histogram corresponding to a reference banknote prepared in advance for each block to identify a banknote.

Next, the processing procedure of the bill validator 10 shown in FIG. 1 will be described.

FIG. 2 is a flowchart showing the processing procedure of the bill validator 10 shown in FIG.

As shown in FIG.
First, the banknote conveyed at high speed is read by the sensor unit 11 to create an input image (step 201).
A smoothing process is performed by applying an integration operator to the created input image to remove noise included in the input image (step 202). The reason for performing the smoothing process is that, when noise is present in the input image, the noise component is emphasized by the second derivative of the noise.

Next, a Laplacian operator is applied to the input image from which noise has been removed to create a Laplacian image corresponding to the input image (step 20).
3) A normalization process for limiting the Laplacian image with predetermined upper and lower limits is performed (step 204) to create a non-linear Laplacian image.

Then, the nonlinear Laplacian image is divided into blocks, a histogram is calculated for each block, and the calculated histogram is compared with a histogram corresponding to a reference bill prepared in advance for each block to identify bills. (Step 205).

As described above, in the bill validating apparatus 10,
Since the pixel value of each pixel of the Laplacian image is suppressed so as not to be equal to or more than the upper limit value and equal to or less than the lower limit value, it is possible to suppress a variation in density distribution and the like caused by a change in the distance between the bill and the image sensor.

Next, the processing concept of the preprocessing unit 12 shown in FIG. 1 will be described.

FIG. 3 is a diagram showing a processing concept of the preprocessing unit 12 shown in FIG. However, here, for convenience of explanation, only the pixel values of the pixels of one line of the input image and the Laplacian image are shown.

FIG. 3A shows the pixel values of one line of the input image. When the input image shown in FIG. 3A is smoothed and the Laplacian operator is applied, FIG.
It becomes as shown in.

As described above, the pixel value of the Laplacian image fluctuates in the + direction or the − direction based on the increase or decrease of the pixel value of the input image. The pixel value is further increased.

For this reason, in this embodiment, FIG.
, Each pixel value of the Laplacian image is set to an upper limit T
In addition to suppressing the value to be equal to or less than 1 and restricting the value to be equal to or less than the lower limit T2.

Therefore, by limiting the Laplacian image with the upper limit value and the lower limit value, the influence of the density fluctuation can be eliminated.

Next, the configuration of the sensor section 11 shown in FIG. 1 will be specifically described.

FIG. 4 is a diagram showing an example of the configuration of the sensor section 11 shown in FIG.

As shown in FIG. 4, the sensor unit 11
The light emitted by the LED 11a is condensed by the rod lens 11b, the banknote 40 conveyed on the conveyance path 41 by the conveyance roller (not shown) is irradiated with light, and the reflected light is detected by the photodiode array 11c. .

Therefore, when the banknote 40 is moved at high speed on the transport path 41, the bill 40 and the photodiode array 1
1c may fluctuate.

When the banknote moves at high speed on the transport path 41, the banknote vibrates and the distance between the sensor and the banknote fluctuates. When the distance becomes longer, the pixel value of the input image decreases and the distance becomes shorter. When this happens, the pixel value of the input image increases. In FIG. 4, the distance between the lower end of the sensor unit 11 and the transport path 41 is about 2 mm.

Although the case where bills are transported on the transport path 41 by the transport rollers has been described here, similar variations in distance and the like also occur when the banknotes are transported using a transport belt or the like.

Next, an operator used by the smoothing processing section 12a and the secondary differentiation processing section 12b shown in FIG. 1 will be described.

FIG. 5A shows the smoothing processing unit 1 shown in FIG.
FIG. 2a is a diagram illustrating an example of an operator used by the operator 2a. The operator illustrated in FIG. 2a is an integration operator that performs smoothing considering eight neighboring pixels.

That is, this operator is a mask pattern for averaging the pixel value of the target pixel located at the center with the pixel values of the eight peripheral pixels, and by applying such a mask pattern, the operator is positioned at the center of the mask. A noise component in which a pixel has a higher peak than a peripheral pixel can be suppressed.

FIG. 5B is a diagram showing an example of an operator used by the second-order differential processing section 12b shown in FIG. 1. The operator shown in FIG. 5 is a representative Laplacian mask having a size of 5 × 5. It is a pattern.

That is, the Laplacian operator is a mask pattern created by replacing the second derivative with a difference in pixel density. In order to apply the Laplacian operator, the Laplacian operator must be within two pixels from the target pixel located at the center. Therefore, the processing of the third line starts when the fifth line is scanned.

The first two lines and the last two lines are excluded from processing because some of the peripheral pixels do not exist.

Although an example of a 5 × 5 Laplacian operator has been described here, a 3 × 3 operator or the like can be used.

Next, a description will be given of a preprocessing result when a test pattern is input to the bill validator 10 shown in FIG.

FIG. 6 is a diagram showing a preprocessing result when a test pattern is input to the bill validator 10 shown in FIG.

FIG. 6A is a diagram showing a test pattern used in place of a banknote. As shown in FIG. 6A, this test pattern is formed by drawing a figure pattern composed of a black-painted rectangle on a horizontally long white paper. These are printed at predetermined intervals.

FIG. 6B is a diagram showing an input image and a nonlinear Laplacian image when this test pattern is read by the image sensor. However, here, for convenience of explanation, only a portion corresponding to the line L1 shown in FIG.

As shown in the figure, when such a test pattern is read four times, the distance between the test pattern and the image data at the time of reading varies, and variations between pixels of the image sensor are present. Even when the same part of the test pattern is read, the density value changes every time the reading is performed.

For this reason, when the Laplacian operator is applied to each of the input images, the variable portion is enlarged.

However, in the present embodiment, the Laplacian value of the pixel affected by the density fluctuation is suppressed by the upper limit value and the lower limit value, and is normalized so as to fall within a certain range. As described above, the nonlinear Laplacian image corresponding to each input image has substantially uniform pixel values.

As described above, by introducing the preprocessing according to the present invention, it is possible to create a nonlinear Laplacian image in which the influence of density fluctuation is eliminated.

As described above, in the present embodiment,
Since the input image read by the sensor unit 11 is smoothed to remove noise components, a Laplacian image is created by applying a Laplacian operator, and the Laplacian image is normalized to create a nonlinear Laplacian image. Of the density distribution caused by the variation of the distance between the image sensor and the image sensor can be suppressed.

In this embodiment, for convenience of explanation,
Secondary differentiation and normalization processing unit 1 performed by secondary differentiation processing unit 12b
Although the normalization processing performed by 2c is described separately, such processing can be performed in parallel, so that high-speed processing can be performed using pipeline processing or the like.

In the present embodiment, the case where the pixel value of each pixel of the Laplacian image is fixedly restricted by using a certain upper limit value and lower limit value has been described, but the present invention is not limited to this. Instead, the pixel value of each pixel of the Laplacian image can be gently limited using a plurality of upper and lower limits.

In the present embodiment, the case where the present invention is applied as a pre-processing unit of the banknote recognition apparatus 10 has been described.
The present invention is not limited to this, and can be applied to various image processing apparatuses.

[0070]

According to the present invention, in order to achieve the above object, if an object during high-speed conveyance is read by an image sensor, the density image read by the image sensor is smoothed and then secondarily differentiated. Since the secondary differential image is formed and each pixel value of the secondary differential image is normalized by limiting it to a predetermined upper limit and lower limit, the following effects can be obtained.

1) It is possible to suppress variations in the density distribution and the like caused by fluctuations in the distance between the bill and the image sensor.

2) The parallel processing of the secondary differentiation and the normalization enables high-speed processing.

[Brief description of the drawings]

FIG. 1 is a diagram showing a configuration of a bill validator used in the present embodiment.

FIG. 2 is a flowchart showing a processing procedure of the bill validator shown in FIG. 1;

FIG. 3 is a view showing a processing concept of a preprocessing unit shown in FIG. 1;

FIG. 4 is a diagram showing an example of a configuration of a sensor unit shown in FIG.

FIG. 5 is a diagram illustrating an example of an operator used by the smoothing processing unit and the secondary differentiation processing unit illustrated in FIG. 1;

FIG. 6 is a view showing a preprocessing result when a test pattern is input to the bill validator shown in FIG. 1;

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 10 ... Banknote identification apparatus, 11 ... Sensor part, 12 ... Preprocessing part, 12a ... Smoothing processing part, 12b ... Secondary differentiation processing part, 12c ... Normalization processing part, 12d ... Threshold value storage part,
13: identification processing unit, 11a: LED, 11b: rod lens, 11c: photodiode array, 40:
Bills, 41 ... Conveyance path

──────────────────────────────────────────────────続 き Continuation of front page (56) References JP-A-7-105365 (JP, A) JP-A-63-54064 (JP, A) JP-A 8-315156 (JP, A) JP-A-5-105 143829 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) G06T 1/00 310 G06T 7/00 300 G07D 7/00

Claims (1)

(57) [Claims] An image processing apparatus for suppressing a variation in density distribution when an object being read at high speed is read by an image sensor, wherein after smoothing the density image read by the image sensor, the density image is subjected to secondary processing. A secondary differential image forming means for differentiating to form a secondary differential image; and, among the pixel values of the secondary differential image formed by the secondary differential image forming means, a pixel value equal to or more than a predetermined upper limit value is set to the upper limit. A normalization unit that normalizes the secondary differential image by replacing the pixel value below a predetermined lower limit value with the lower limit value, based on the secondary differential image normalized by the normalization unit. An image processing apparatus comprising: an identification unit that identifies the object.