JPH01271885A - Direction curvature extracting system for curved striped pattern - Google Patents

Abstract

PURPOSE:To extract the curvature of a curved striped pattern and furthermore to accurately extract the pattern direction by accumulating the variance of the gradation value in each quantizing direction and deciding the quantizing direction and the curvature from the extreme accumulation value. CONSTITUTION:A picture memory part 10 stores the gradation value of a striped pattern quantized into the picture element formed in a two-dimensional array shape. A direction curvature memory part 20 stores the direction curvature of an extracted striped pattern and a control part 30 produces various types of control signals. An address converting part 40 designates the address of the gradation value to be read out by the part 10. A variance calculating part 50 calculates the absolute value of the difference among gradation values of the continuous picture elements. An arc-shaped direction deciding part 60 accumulates the values given from the part 50 for each quantizing arc-shaped direction and also calculates the extreme accumulation value. When said extreme value is obtained at the part 60, the arc-shaped direction value is given to the part 20 when the processing is through to a single center picture element.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a curved striped pattern direction curvature extraction method used when extracting striped patterns such as fingerprints and fish scales.

[Conventional technology]

Fingerprints, which are generally made up of many ridges in a single curved striped pattern, have two major characteristics: they remain unchanged throughout life and are the same for everyone, so they have long been used as a means of identifying people in criminal investigations. . And in recent years,
A system has been realized that automatically performs this fingerprint verification process using computer-based pattern recognition technology.

In such systems, minutiae are scattered on the fingerprint.

For example, the fingerprint matching process is automated by extracting end points where ridges break and branching points where ridges diverge and merge, and comparing these feature points. Therefore, it is extremely important to extract the positions of minutiae from fingerprints.

On the other hand, in a striped pattern made up of many ridges like a fingerprint, the flowing direction and curvature of the ridges are very stable information, so if the curvature in the ridge direction can be extracted, the matching and identification process will be smooth. And it can be done stably.

Conventionally, as a method for extracting the direction of a striped pattern, the direction of the stripes in a ruru pixel of a striped pattern image is based on the fact that the variation in shading is small in the same direction as the stripes, and the variation is large in the direction perpendicular to the stripes. There is a method of finding the extreme value of a plurality of predetermined variations in the density of quantization power improvement, and determining the direction of the stripes from this extreme value (for example, Japanese Patent Publication No. 52-97298
Publication No.).

[Problem to be solved by the invention]

In the above-mentioned conventional extraction method, a striped pattern gradation image is linearly traced to obtain variations in a-darkness.

Therefore, in principle, the curvature of the striped pattern cannot be extracted. In addition, in the conventional extraction method, since the striped pattern gradation image is traced linearly, areas where the ridge curvature is large, such as the center of the fingerprint and the vicinity of the triangular part, are shown in Figures 5 (a) to (c). Even when tracing is done in the same direction as the ridge, there is a problem that the shading varies considerably and the reliability of extracting the ridge direction is low. That is, even if traced in the direction of the ridge, since the curvature is large, the ridge (hatched area) intersects the valley (white area), resulting in large variations in shading.

[Means to solve the problem]

The method for extracting the directional curvature of a curved striped pattern according to the present invention includes: an image storage unit that holds the grayscale values of image pixels of a curved striped pattern grayscale image quantized into a two-dimensional array of pixels; Specify the address of an image pixel stored in the storage unit, and create an image that is located in a radial straight line or a radial curve with a predetermined quantization power improvement, centered on the position of the specified image pixel. an address converting section that sequentially specifies elementary addresses, and an image that receives grayscale values of image pixels from addresses of the image storage section specified by the address converting section and sequentially reads out in a radial straight line or curved shape. a fluctuation calculation unit that calculates an increasing function value of the absolute value of the difference between the gray value of a picture element and a previously read gray value; and a fluctuation calculation unit that calculates a cumulative sum of the fluctuation amount that is the output of the fluctuation calculation unit for each of the quantization directions. The device is characterized by having a direction curvature determining section that determines multiple quantization directions and curvature at the extreme value of the cumulative sum. It also becomes possible to extract curvature.

〔Example〕

The present invention will be explained below with reference to the drawings.

First, the principle of the present invention will be explained with reference to FIGS. 1(a) and 1(b). Note that FIG. 1(a) shows the quantized 2
Figure 1(b) shows the dimensional striped pattern bridge light image, and Figure 1(b) is the same as Figure 1(a).
) is a partially enlarged diagram.

Referring to FIGS. 1(,) and (b), the striped pattern represented by the fingerprint image has a plurality of layered ridges f, and the direction and curvature thereof are very stable information. This partial image shown in FIG. 1(b) shows a ridge f flowing in a curved line in one direction.

Here, the ridge f in the same direction (d) as the tangential direction of the ridge f
The arcuate direction with the same curvature (C) as Ad, c
(This direction is hereinafter defined as the arcuate direction).

A case will be described in which the center picture element is Vm and a white background portion is selected. Considering the density variation of the pixel density for the pixel rows Ld, c arranged in the same arcuate direction Ad, c as the ridge f, the density of the central pixel Vm and the central pixel V
Picture elements Ld, c', which are separated from m by a quantization distance r1,
The densities of 1 are equal to each other when there is no dirt or the like on the paper surface, and therefore no variation in density occurs. Similarly, the pixel densities Ld, c, 2 to Ld, c, n at points distant from the center pixel Vm by distances r2, . , 2~Ld
, c; When compared with n-1, the density variation is substantially small.

Here, with a certain picture element on the two images as the center, and this painting as the center painting Vm, the quantization distance r1 from the center painting Vm,
...The operation of comparing the continuous pixel hardness at points separated by rn is carried out in various arcuate directions with one curvature, and the added value of the density difference is calculated for each direction, and the added values of each arcuate direction are compared. It can be seen that the added value in the arcuate directions Ad and c of the ridge f is minimal. Therefore, the central picture element V
In the vicinity of m.

By performing the operation to find the minimum value described above.

The directional curvature of the ridge if can be extracted.

FIG. 2(a) shows fluctuations in picture element appetite. Figure 2(a)
Here, the horizontal axis represents the quantization distance r from the central picture element Vm, and the vertical axis represents the picture element distance p. As shown in FIG. 2(a), picture element rows Ld arranged in the same arcuate directions Ad and c as the ridge f
, c are extremely small over the distances r1 to rn, but in directions other than the arcuate direction of the ridge if, for example.

Picture element rows Ld', c' arranged in arcuate directions Ad', c'
The variation in brightness is much larger than that of the picture element rows Ld and c.

In Fig. 2(b), the distance r from the central picture element Vm is plotted on a different axis, and the number of integrated values G (r) of the variation in the number of pixels of the picture element rows Ld, c, Ld', c' is plotted on the vertical axis. The fluctuations in pixel density taken during the period are shown in b.

Here, if the variation in pixel density is f (r) and a predetermined monotonically increasing function g is selected, the above-mentioned integral value can be expressed in discrete form as G(rn)-Σg (l f (r,)-f ( r,
, ) l ).

As shown in FIG. 2(b), the number of integrated values Gd, ber) for the picture element rows Ld, c increases rapidly (((()) for the picture element row Ld' + c' due to variations in shading. Since the number of integral values Cd', c' (r) passes through a path with small fluctuations of − degrees, the rate of increase is smaller than that of Gd, c (r).For this reason, the number of integral values Cd', c' (r) is smaller than that of Gd, c' (r).
Comparing G (r) at positions separated by Gd
, c(R) is smaller than Gd', c'(R). Therefore, a predetermined arcuate direction d i+
CJ (i-1+ 2+...l(, j=1.2.-
....

Let G(r) at t) be c ] I j(r) and find Gi and j(R) at a constant distance R, respectively.
The arcuate direction d1 in the central picture element Vm. Cjba.

dc-d・, C1r′″1nG1.j(R)1
Ko I J ],
It can be expressed as

In this way, 1! Since the directional curvature of the ridge is determined by calculating the absolute value of the light variation difference and integrating this, even if there is dirt or the like that causes noise in the path.

Malfunctions do not occur due to these influences. Therefore.

The directional curvature extraction operation can be performed very stably and smoothly.

Next, two aspects of the present invention will be specifically explained with reference to FIG.

The directional curvature extraction method according to the present invention includes an image storage unit 1 that stores striped pattern gray values quantized into a two-dimensional array of picture elements.
0. A directional curvature storage unit 20 that stores the directional curvature of the extracted striped pattern. and a control section 30 that generates various control signals, and an address conversion section 40 that specifies the address of the gray value to be read out from the one-image storage section 10. Fluctuation calculation unit 5 that calculates the insulation value of the difference in gray values of consecutive picture elements
0, and an arcuate direction determination unit 60 that calculates the cumulative sum of the values from the fluctuation calculation unit 50 for each quantized arcuate direction and calculates the minimum value of the cumulative sum. When the minimum value is found in the arcuate direction determining section 60, the arcuate direction value is given to the direction curvature storage section 20 after completing the processing for one central picture element.

On the other hand, the control unit 30 controls the address of the center picture element of the image storage unit 10. , Jo, and this address. , J.
To read the gray value near o.

Specify the quantization arc direction d i r (! j and the quantization distance r from the center picture element. Next, the address conversion unit 40
is the address ■ of the center picture element from the control unit 30. , Jo, performs address conversion and specifies the corresponding address in the one-image memory 30, and furthermore, the arcuate direction d i is sent from the control unit 3o.
+Cj and distance rK, specify an address in the specified arcuate direction d i , (!J and distance rK from the center picture element. To perform such an operation, the arcuate direction d1. The difference in addresses from the center picture element is stored in advance according to the distance rK.

The above-mentioned address difference may be added to the address of the center picture element at the time when the arcuate direction d i r Cj and the distance stroke are specified.

Let's take the case of finding the directional curvature of a fingerprint ridge as an example.
explain.

Here, as one direction curvature of a fingerprint ridge, eight kinds of directions shown in FIG.
Assume five types of curvature shown in Figure (b). Therefore 2
A total of 40 types of arcuate directions are possible.

When extracting the curvature in one direction of an arbitrary central picture element.

That address I. , Jo and the first arc direction d. 1 eO
and the initial distance r. is given to the address converter 40, and the corresponding coordinates are determined. Based on these coordinates, the density of the picture element is extracted from the image storage section 10 and sent to the variation calculation section 50.

Then the same arcuate direction d. + eo, the density for the distance r1 is similarly sent to the variation calculation unit 50, and the absolute value of the density difference is integrated. The above process is repeated until the distance rn is completed. arcuate direction d. ,c.

The final amount of variation in is determined by the arcuate direction determination unit 60
sent to.

Next, the arcuate direction d. , el are calculated in the same way, and sequentially in the last arcuate direction d7. The amount of variation in each arcuate direction is calculated up to c4. When the amount of variation is determined for all arc directions, the minimum value thereof is determined by the arc direction determination unit 60, and the arc direction di + (!j, that is, the tangential direction di, and the curvature cj is the directional curvature. It is registered in the storage unit 20.

In this way, by calculating the amount of variation using an arcuate direction with one curvature in the direction of the glaze, it is possible to find an arcuate direction with extremely small amount of variation even in areas where the ridge curvature is large. As a result, the direction and curvature of the ridge are accurately extracted.

That is, in the case of a ridge with a large curvature as shown in FIG. 5(b), if it is traced in the arcuate direction Ld 1 * c 3,
There is almost no variation in shading, and the direction and curvature can be determined accurately. However, when tracing is performed in a linear direction as in the conventional case, even if the tracing direction is the same as the ridge (Ldj, co).

It intersects with other ridges, there are large variations in shading, and it is difficult to extract the direction.

〔Effect of the invention〕

As explained above, according to the present invention, it is possible to extract the 8-curved curvature of the number of stripes, and it is also possible to accurately extract the pattern direction.

[Brief explanation of the drawing]

FIGS. 1(a) and (b) are diagrams for explaining the extraction of curvature in two directions, FIG. 2 is a diagram for explaining the present invention in detail,
FIG. 3 is a diagram for explaining the state of a striped pattern from which one curvature in the block direction is extracted, showing an embodiment of the present invention. 10: Image memory, 20: Direction memory, 30: Control unit, 40
Near address conversion section, 50: Fluctuation calculation section. 60: Arc-shaped direction determining section. Figure 1 ゛(Q) (b) 2nd (Q) G(r) (b)

Claims (1)

[Claims] An image memory that is used when extracting the pattern direction and curvature of a curved striped pattern gradation image quantized into a one- and two-dimensional array of pixels, and holds the gradation values of the pixels of the quantized image. area and the address of the image pixel stored in the image storage unit, and the image is created in a radial straight line or a radial curve shape with a predetermined quantization power improvement centered on the position of the specified image pixel. an address conversion section that sequentially specifies the addresses of image pixels located at the address conversion section; and an image picture that receives the grayscale values of the image pixels at the addresses specified by the address conversion section and sequentially reads them out in a radial straight line or a curved line. a fluctuation calculation unit that calculates an increasing function value of the absolute value of the difference between the original gray value and the previously read gray value, and cumulatively summing the fluctuation amount from the fluctuation calculation unit for each of the quantization directions, A directional curvature extraction method for a curved striped pattern, comprising a directional curvature determining unit that determines a quantization direction and a curvature based on the extreme value of the cumulative sum.