JPH0520443A - Registration method for fingerprint data - Google Patents

Abstract

PURPOSE:To hold information on an original picture while a manutia characteristic is clearly shown by thinning a fingerprint and registering the binary pattern of eight peripheral picture elements in a 3X3 convolution in which a center picture is a graphic picture element as the density of the center picture element. CONSTITUTION:Noise removal 1-2, optimum binarization 1-3 and thinning 1-4 are sequentially executed by following fingerprint picture fetch 1-1. Data generation/registration 1-5 using a lookup table is executed. The lookup table converts the thinning pattern of the 3X3 convolution into numerical data, and data on the respective picture elements after conversion are registered in an image memory. Since the thinning pattern whose center picture element is the graphic becomes an evaluation object, all the patterns are expressed by the pattern of eight peripheral picture elements. A bit string is generated by giving '0' (right of the center picture element) to '7' anticlockwise to the peripheral picture elements.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for registering fingerprint data in fingerprint collation.

[0002]

2. Description of the Related Art As a fingerprint collation method, a minutia network method using end points and bifurcation points as feature quantities is highly reliable, backed by many years of experience. However, it is difficult to prevent noise from being mixed in when inputting a fingerprint image, and it is undeniable that not only data storage by minutia but also image data storage is necessary. That is, in the conventional fingerprint collation, the information of the original image data is largely lost by extracting only the minutiae, and the necessity of storing the original image is high.

[0003]

SUMMARY OF THE INVENTION The present invention was devised to solve such conventional problems, and provides a data registration method capable of clearly showing minutiae features while preserving information of an original image as much as possible. The purpose is to

[0004]

According to the fingerprint data registration method of the present invention, a fingerprint image is thinned, pixels in each part of the image in this thinned image are converted into numerical data corresponding to the pattern of the part, and the numerical data Is registered as the pixel value of the pixel. According to the fingerprint data registration method according to the present invention, minutia features can be specified while preserving the thinning pattern of each part of the image.

[0005]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of a fingerprint data registration method according to the present invention will be described below with reference to the drawings. FIG. 1 is a flowchart showing a series of processes from fingerprint image capture to data registration. Following fingerprint image capture (step 1-1), noise removal (step 1-2) and optimal binarization (step 1-). 3), thinning (step 1-4) is performed sequentially, and finally data generation using a lookup table,
Register. The lookup table converts the thinning pattern of 3 × 3 convolution into numerical data, and the data of each pixel after conversion is registered in the image memory.

Since the thinning pattern is evaluated only when the central pixel is a figure, the entire pattern is expressed only by the pattern of the eight surrounding pixels. In FIG. 2, the peripheral pixels are numbered from "0" (right of the center pixel) to "7" counterclockwise.
The bit string is generated with the number LSB and the number "7" as the MSB.

The 8-bit pattern corresponds to each minutiae, and the pattern can be directly associated with the minutiae. Further, the characteristic of a simple curve is determined by the surrounding two pixels. The relationship between each minutiae and bit pattern (hexadecimal notation) is shown below.

End point 01, 02, 04, 08, 10, 20, 40, 80 Branch point 15, 25, 29, 2A, 45, 49, 4A 51, 52, 54, 8A, 92, 94, A2, A8 Straight line 11, 22, 44, 88 Curve 03, 05, 06, 09, 0A, 0C 12, 14, 18, 21, 28 30, 41, 42, 44, 48, 50, 60 81, 82, 84, 90, A0 , C0

As an example of these patterns, the end point “02” is the pattern of FIG. 3, the curve “0A” is the pattern of FIG. 4, the branch point “2A” is the pattern of FIG. 5, and the straight line “2”.
2 "becomes the pattern of FIG. In this way, if the surrounding 8 pixels of the 3 × 3 convolution in the thinned image are treated as 8-bit data and registered as the density of the central pixel in the image memory, the minutiae can be easily extracted and the thinned image can be extracted. Can be completely reproduced.

[0010]

As described above, in the fingerprint data registration method according to the present invention, the fingerprint image is thinned, the pixels of each part of the image in the thinned image are converted into numerical data corresponding to the pattern of the part, and Since the numerical data is registered as the pixel value of the pixel, it has an excellent effect that the minutia feature can be easily specified while preserving the thinning pattern of each part of the image.

[Brief description of drawings]

FIG. 1 is a flowchart showing an embodiment of a fingerprint data registration method according to the present invention.

FIG. 2 is a conceptual diagram for defining surrounding pixels in 3 × 3 convolution.

FIG. 3 is a conceptual diagram showing an end point pattern.

FIG. 4 is a conceptual diagram showing a curve pattern.

FIG. 5 is a conceptual diagram showing a pattern of branch points.

FIG. 6 is a conceptual diagram showing a linear pattern.

Claims (1)

Claim: What is claimed is: 1. A fingerprint image is thinned, and in this thinned image, a binary pattern of 8 peripheral pixels in a 3x3 convolution in which the central pixel is a graphic pixel is registered as the density of the central pixel. A method for registering fingerprint data, characterized by.