Methods and apparatus for estimating orientation in an image

• US 20070047785A1
• Filed: 08/23/2006
• Published: 03/01/2007
• Est. Priority Date: 08/23/2005
• Status: Active Grant

• Current Assignee: Samsung Electronics Co. Ltd.
• Original Assignee: Samsung Electronics Co. Ltd.
• Inventors: Lee, Dong-jae, Jang, Won-churl, Park, Deok-soo
• Granted Patent: US 7,876,933 B2
• Time in Patent Office: Days
• US Class Current: 382/125
• CPC Class Codes: G06T 2207/30004 Biomedical image processing
• CPC Class Codes: G06T 7/73 using feature-based methods
• CPC Class Codes: G06T 7/77 using statistical methods
• CPC Class Codes: G06V 10/242 by image rotation, e.g. by 90 degrees
• CPC Class Codes: G06V 40/1359 Extracting features related to ridge properties; Determining the fingerprint type, e.g. whorl or loop

Abstract

Methods and apparatus for estimating an orientation are disclosed. Methods of estimating an orientation include: dividing a fingerprint image into first partial regions, measuring the gradient of each pixel of the fingerprint image, and estimating a representative orientation of each of the first partial regions; obtaining an improved fingerprint image by filtering the fingerprint image using a double orientation filter, and remeasuring the representative orientation of each of the first partial regions by remeasuring the gradient of each pixel of the improved fingerprint image; and dividing the first partial regions into second partial regions, and estimating the representative orientation of the second partial regions with respect to the curvatures of the first partial regions in response to the remeasured representative orientation of each of the first partial regions and the remeasured gradients of the pixels.

Claims

1. A method of estimating an image orientation field, comprising:

dividing an image into first partial regions;

measuring gradients of pixels of the image;

estimating representative orientations of the first partial regions;

filtering the image using a double orientation filter to form an improved image;

remeasuring the representative orientations of the first partial regions by remeasuring the gradients of the pixels of the improved image;

dividing the first partial regions into second partial regions; and

estimating representative orientations of the second partial regions with respect to curvature of the first partial regions in response to the remeasured representative orientations of the first partial regions and the remeasured gradients of the pixels.

2. The method of claim 1, wherein estimating the representative orientations of the first partial regions comprises:

dividing the image into the first partial regions;

measuring the gradients of pixels of the first partial regions with respect to a first orientation and a second orientation using gradient measuring operators in the first orientation and the second orientation, wherein the gradient measuring operators are perpendicular to each other; and

measuring the orientations of pixels of the first partial regions using the measured gradients of the pixels and measuring the representative orientations of the first partial regions using a histogram with respect to the measured orientations of pixels of the first partial regions.

3. The method of claim 2, wherein in the gradient measuring operator of the first orientation, a pixel whose gradient is measured is taken as a central pixel, and weights of pixels immediately adjacent to the central pixel are less than weights of pixels second-adjacent to the central pixel in the first orientation, and

in the gradient measuring operator of the second orientation, a pixel whose gradient is measured is taken as a central pixel, and weights of pixels immediately adjacent to the central pixel are less than weights of pixels second-adjacent to the central pixel in the second orientation.

4. The method of claim 3, wherein the first orientation is the x orientation in a rectangular coordinate system and the second orientation is the y orientation in the rectangular coordinates system,

the gradient measuring operator (Ox(x,y)) in the first orientation has a size of 5×

5 pixels and the coordinates of the central pixel are (2,

     1), values in coordinates of the gradient measuring operator in the first orientation are Ox(0,

     0)=1, Ox(1,

     0)=0, Ox(2,

     0)=0, Ox(3,

     0)=0, Ox(4,

     0)=−

1, Ox(0,

     1)=2, Ox(1,

     1)=1, Ox(2,

     1)=0, Ox(3,

     1)=−

1, Ox(4,

     1)=−

2, Ox(0,

     2)=1, Ox(1,

     2)=0, Ox(2,

     2)=0, Ox(3,

     2)=0, Ox(4,

     2)=−

1, and

the gradient measuring operator (Oy(x,y)) in the second orientation has a size of 3×

5 pixels and the coordinates of the central pixel are (1,

     2), values in coordinates of the gradient measuring operator in the second orientation are Oy(0,

     0)=−

1, Oy(0,

     1)=0, Oy(0,

     2)=0, Oy(0,

     3)=0, Oy(0,

     4)=1, Oy(1,

     0)=−

2, Oy(1,

     1)=−

1, Oy(1,

     2)=0, Oy(1,

     3)=1, Oy(1,

     4)=2, Oy(2,

     0)=−

1, Oy(2,

     1)=0, Oy(2,

     2)=0, Oy(2,

     3)=0, Oy(2,

     4)=1.

5. The method of claim 2, wherein measuring the gradients of pixels of the image comprises measuring the gradients of the first orientation and the second orientation by performing convolution operations of gradient measuring operators of the first orientation and the second orientation with pixels neighboring a pixel whose gradient is being measured.

6. The method of claim 5, wherein the gradient in the first orientation is measured using the following equation:

$\mathrm{xGradient}<br><br>\left(x,y\right)=\underset{i=-2}{\overset{2}{\sum <br><br>}}<br><br>\underset{j=-1}{\overset{1}{\sum <br><br>}}<br><br>p<br><br>\left(x+i,y+j\right)<br><br>O_x<br><br>\left(i+2,j+1\right)$

and the gradient in the second orientation is measured using the following equation;

$\mathrm{yGradient}<br><br>\left(x,y\right)=\underset{i=-1}{\overset{1}{\sum <br><br>}}<br><br>\underset{j=-2}{\overset{2}{\sum <br><br>}}<br><br>p<br><br>\left(x+i,y+j\right)<br><br>O_y<br><br>\left(i+1,j+2\right)$

where xGradient(x, y) and yGradient(x, y) denote the first orientation gradient and the second orientation gradient, respectively, of a pixel at coordinates (x, y), p(x, y) denotes a grayscale value of a pixel at coordinates (x, y), Ox(x, y) and Oy(x, y) denote the first gradient measuring operator and the second gradient measuring operator, respectively with a pixel at coordinates (x, y) as a central pixel, and the first orientation gradient measuring operator and the second orientation gradient measuring operator have sizes of 5×

5 pixels and 3×

5 pixels, respectively.

7. The method of claim 2, wherein the orientation of pixels is measured in response to a ratio of the first orientation gradient and the second orientation gradient.

8. The method of claim 7, wherein the orientation of pixels is measured using the following equation:

$O<br><br>\left(x,y\right)=\frac{180}{\mathrm{\pi <br><br>}}<br><br>{\tan }^{-1}<br><br>\left(\frac{\mathrm{yGradient}<br><br>\left(x,y\right)}{\mathrm{xGradient}<br><br>\left(x,y\right)}\right)$

where O(x, y) denotes the orientation of a pixel at coordinates (x, y), and Gx(x, y) and Gy(x, y) denote the first and second orientation gradients, respectively, of a pixel at coordinates (x, y).

9. The method of claim 8, wherein the histogram is obtained by quantizing the orientations in predetermined angular units.

10. The method of claim 9, wherein the histogram is expressed as the following equation:

$H_{\mathrm{MA}}<br><br>\left(O<br><br>\left(x,y\right)/Z\right)=\underset{i=-M}{\overset{M}{\sum <br><br>}}<br><br>\underset{l=M}{\overset{M}{\sum <br><br>}}<br><br>K(O<br><br>\left(\mathrm{mN}+i,\mathrm{nN}+j\right)/Z$

where Hmn denotes the histogram of a first partial region at a location which is an m-th partial region in the x orientation and an n-th partial region in the y orientation, and K(1)=1, and 1 is a value obtained by quantizing the orientation of pixel.

11. The method of claim 2, wherein an orientation corresponding to a location in which the histogram has a maximum value is taken as the representative orientation of a first partial region.

12. The method of claim 2, wherein measuring the representative orientation of first partial regions further comprises adjusting a brightness uniformity of the image by processing a brightness characteristic of the first partial regions, using a brightness characteristic of the image.

13. The method of claim 12, wherein the adjusting the image comprises:

measuring an image uniformity degree in response to an absolute value of a brightness difference among pixels in a first partial region;

determining that the fingerprint pattern of the region is uniform if the fingerprint pattern uniformity degree is less than a predetermined uniformity degree threshold, or else determining that the fingerprint pattern of the region is not uniform; and

making the fingerprint pattern uniform by low-pass filtering the region determined to be non-uniform.

14. The method of claim 13, wherein the fingerprint pattern uniformity degree is measured using the following equation:

$V_{\mathrm{noise}}<br><br>\left(m,n\right)=\underset{i=0}{\overset{N}{\sum <br><br>}}<br><br>\underset{j=0}{\overset{N}{\sum <br><br>}}<br><br><br><br>x_{\left(\mathrm{mN}+i,\mathrm{nN}+j\right)}-x_{\left(\mathrm{mN}+i-1,\mathrm{nN}+j\right)}<br><br>$

where Vnoise(m, n) denotes the fingerprint pattern uniformity degree of a first partial region at a location which is an m-th region in the x orientation and an n-th region in the y orientation, and x(x, y) denotes the brightness of a pixel at coordinates (x, y).

15. The method of claim 13, wherein the low-pass filtering is performed using a Gaussian filter.

16. The method of claim 1, wherein the image comprises a fingerprint image, and wherein remeasuring the representative orientation comprises:

determining whether the image of a first partial region is clear or not by measuring coefficients indicating whether a fingerprint ridge and a fingerprint valley are clearly distinguished or not;

filtering images of the first partial regions using a double orientation filter in response to whether the image is determined to be clear or not; and

remeasuring the representative orientations of the first partial regions by remeasuring the gradient of pixels of the filtered image.

17. The method of claim 16, wherein determining whether the image is clear or not comprises:

measuring a distinction coefficient indicating the degree that the ridges and valleys of the image are clearly distinguished, and a single orientation coefficient indicating the degree that the structure of the ridge and valley of the image form a single orientation; and

if the distinction coefficient is greater than a predetermined distinction threshold and the single orientation coefficient is greater than a predetermined single orientation threshold, determining that the image is clear, and if the distinction coefficient is not greater than the predetermined distinction threshold or the single orientation coefficient is not greater than the predetermined single orientation threshold, determining that the image is not clear.

18. The method of claim 17, wherein the distinction coefficient is measured using the following equation:

$C_{h<br><br>\text{\hspace{1em}<br><br>}<br><br>1}=\frac{\begin{array}{c}\underset{x=0}{\overset{N}{\sum <br><br>}}<br><br>\underset{y=0}{\overset{N}{\sum <br><br>}}<br><br>\mathrm{xGradient}<br><br>\left[y\right]<br><br>\left[x\right]·<br><br>\mathrm{xGradient}<br><br>\left[y\right]<br><br>\left[x\right]+\\ \underset{x=0}{\overset{N}{\sum <br><br>}}<br><br>\underset{y=0}{\overset{N}{\sum <br><br>}}<br><br>\mathrm{yGradient}<br><br>\left[y\right]<br><br>\left[x\right]·<br><br>\mathrm{yGradient}<br><br>\left[y\right]<br><br>\left[x\right]\end{array}}{N·<br><br>N}.$

and the single orientation coefficient is measured using the following equation;

$C_{h<br><br>\text{\hspace{1em}<br><br>}<br><br>2}=\frac{\underset{y=0}{\overset{N}{\sum <br><br>}}<br><br>\underset{x=0}{\overset{N}{\sum <br><br>}}<br><br>{\left\{\mathrm{xGradient}<br><br>\left[y\right]<br><br>\left[x\right]·<br><br>\cos <br><br>\text{\hspace{1em}<br><br>}<br><br>\mathrm{\theta <br><br>}+\mathrm{yGradient}<br><br>\left[y\right]<br><br>\left[x\right]·<br><br>\sin <br><br>\text{\hspace{1em}<br><br>}<br><br>\mathrm{\theta <br><br>}-{\stackrel{\_}{m}}_{\stackrel{\_}{y}}\right\}}^2}{\underset{y=0}{\overset{N}{\sum <br><br>}}<br><br>\underset{x=0}{\overset{N}{\sum <br><br>}}<br><br>{\left\{\mathrm{yGradient}<br><br>\left[y\right]<br><br>\left[x\right]·<br><br>\cos <br><br>\text{\hspace{1em}<br><br>}<br><br>\mathrm{\theta <br><br>}-\mathrm{xGradient}<br><br>\left[y\right]<br><br>\left[x\right]·<br><br>\sin <br><br>\text{\hspace{1em}<br><br>}<br><br>\mathrm{\theta <br><br>}-{\stackrel{\_}{m}}_{\stackrel{\_}{x}}\right\}}^2}$ ${\stackrel{\_}{m}}_{\stackrel{\_}{y}}=\frac{\underset{x=0}{\overset{N}{\sum <br><br>}}<br><br>\underset{y=0}{\overset{N}{\sum <br><br>}}<br><br>\mathrm{xGradient}<br><br>\left[y\right]<br><br>\left[x\right]·<br><br>\cos <br><br>\text{\hspace{1em}<br><br>}<br><br>\mathrm{\theta <br><br>}+\mathrm{yGradient}<br><br>\left[y\right]<br><br>\left[x\right]·<br><br>\sin <br><br>\text{\hspace{1em}<br><br>}<br><br>\mathrm{\theta <br><br>}}{N·<br><br>N}$ ${\stackrel{\_}{m}}_{\stackrel{\_}{x}}=\frac{\underset{x=0}{\overset{N}{\sum <br><br>}}<br><br>\underset{y=0}{\overset{N}{\sum <br><br>}}<br><br>\mathrm{xGradient}<br><br>\left[y\right]<br><br>\left[x\right]·<br><br>\cos <br><br>\text{\hspace{1em}<br><br>}<br><br>\mathrm{\theta <br><br>}+\mathrm{xGradient}<br><br>\left[y\right]<br><br>\left[x\right]·<br><br>\sin <br><br>\text{\hspace{1em}<br><br>}<br><br>\mathrm{\theta <br><br>}}{N·<br><br>N}$

where Ch1 denotes the distinction coefficient and Ch2 denotes the single orientation coefficient, and the distinction coefficient indicates whether or not the gradients of the first and second orientations have a concentrated distribution and the single orientation coefficient indicates whether the gradients of the first and second orientations are distributed in a circular shape and/or in an elliptical shape.

19. The method of claim 16, wherein the improved image is obtained by removing noise existing in a ridge and a valley in the first partial regions, by using a first orientation filter emphasizing the boundary of a ridge and a valley in relation to four predetermined orientations, and a second orientation filter removing noise existing inside the ridge and valley in relation to the four orientations, and

wherein the double orientation filter includes the first and second orientation filters.

20. The method of claim 19, wherein in the first orientation filter, a pixel in which a boundary of a ridge and a valley is desired to be emphasized is taken as a central pixel, and in relation to each of the four predetermined orientations, pixels immediately adjacent to the central pixel have the same weight as the weight of the central pixel, and pixels second-adjacent to the central pixel have weights whose sign is opposite to the sign of the weight of the central pixel and whose absolute value is less than the absolute value of the weight of the central pixel, and pixels third-adjacent to the central pixel have weights whose sign is opposite to the sign of the central pixel and whose absolute value is the same as the absolute value of the weight of the central pixel.

21. The method of claim 20, wherein in the first orientation filter, the weight of the central pixel and the weight of pixels immediately adjacent to the central pixel are 2 each, the weight of pixels second-adjacent to the central pixel is −

1, and the weight of pixels third-adjacent to the central pixel is −

2.

22. The method of claim 19, wherein in the second orientation filter, a pixel in which noise inside a ridge and a valley is desired to be removed is taken as a central pixel, and the central pixel has a maximum weight, and pixels immediately adjacent to the central pixel and pixels second-adjacent to the central pixel have weights less than the weight of the central pixel, and pixels third-adjacent to the central pixel have weights less than the weight of pixels second-adjacent to the central pixel.

23. The method of claim 22, wherein in the first orientation filter, the weight in the central pixel is 3, the weight of pixels immediately adjacent to the central pixel and the weight of pixels second-adjacent to the central pixel are 2, and the weight of pixels third-adjacent to the central pixel is 1.

24. The method of claim 19, wherein obtaining the improved image comprises:

filtering the first partial regions that are determined to be clear with a first orientation filter and a second orientation filter for an orientation corresponding to the representative orientation of the first partial region; and

measuring clearness in four orientations of first partial regions that are determined to be unclear using the first orientation filter and the second orientation filter for each of the four orientations, filtering the first partial regions that are determined to be unclear with the first orientation filter and the second orientation filter in an orientation in which the clearness is a maximum, measuring the clearness in each orientation of the first partial region in response to the power of each orientation of the first partial region after filtering using the first orientation filter, and measuring the single orientation coefficient of each orientation of the first partial region after filtering using the second orientation filter.

25. The method of claim 24, wherein clearness is measured using the following equation:

$C_{\mathrm{reactitvity}}=\mathrm{\alpha <br><br>}·<br><br>C_{{\mathrm{RS}}_l}+\mathrm{\beta <br><br>}·<br><br>C_{{\mathrm{RNB}}_i}$ $C_{{\mathrm{RS}}_i}=100·<br><br>\frac{\underset{x=0}{\overset{N}{\sum <br><br>}}<br><br>\underset{y=0}{\overset{N}{\sum <br><br>}}<br><br>\left\{\mathrm{Inimage}<br><br>\left[y\right]<br><br>\left[x\right]\otimes <br><br>{\mathrm{RSF}}_i\right\}}{{\mathrm{MaxPower}}_o}$ $C_{{\mathrm{RNR}}_i}=\frac{N<br><br>\underset{y=0}{\overset{N}{\sum <br><br>}}<br><br>\underset{x=0}{\overset{N}{\sum <br><br>}}<br><br>{\left\{\mathrm{xGradient}<br><br>\left[y\right]<br><br>\left[x\right]·<br><br>\cos <br><br>\text{\hspace{1em}<br><br>}<br><br>\mathrm{\theta <br><br>}+\mathrm{yGradient}<br><br>\left[y\right]<br><br>\left[x\right]·<br><br>\sin <br><br>\text{\hspace{1em}<br><br>}<br><br>\mathrm{\theta <br><br>}-{\stackrel{\_}{m}}_{\stackrel{\_}{y}}\right\}}^2}{\underset{y=0}{\overset{N}{\sum <br><br>}}<br><br>\underset{x=0}{\overset{N}{\sum <br><br>}}<br><br>{\left\{\mathrm{yGradient}<br><br>\left[y\right]<br><br>\left[x\right]·<br><br>\cos <br><br>\text{\hspace{1em}<br><br>}<br><br>\mathrm{\theta <br><br>}-\mathrm{xGradient}<br><br>\left[y\right]<br><br>\left[x\right]·<br><br>\sin <br><br>\text{\hspace{1em}<br><br>}<br><br>\mathrm{\theta <br><br>}-{\stackrel{\_}{m}}_{\stackrel{\_}{x}}\right\}}^2}$

where Creactivity i denotes the clearness in i-th orientation in each first partial region, CRSi denotes the power in i-th orientation in each first partial region after filtering using the first orientation filter, CRNRi denotes the single orientation coefficient in i-th orientation of each first partial region after filtering using the second orientation filter, and a and fi denote predetermined constants corresponding to conditions under which the image is taken.

26. The method of claim 1, further comprising, after remeasuring the representative orientation, measuring the entire orientation of adjacent first partial regions and correcting an error of the remeasured representative orientation of the first partial regions in response to a difference of the entire orientation and the remeasured representative orientations of the first partial regions.

27. The method of claim 26, wherein correcting the error of the first partial regions comprises:

measuring the entire orientation of adjacent first partial regions in response to the mean value of a difference of the remeasured representative orientation and a predetermined angle;

determining whether or not an error occurs in response to the difference of the measured entire orientation and the remeasured representative orientation of the first partial regions; and

for first partial regions in which an error occurs, if there is a vertical orientation identity or a horizontal orientation identity with the adjacent first partial regions above, below, to the left or to the right of the first partial region, correcting the error in response to the mean value of the adjacent first partial regions in the horizontal direction, and if there is no vertical orientation identity or horizontal orientation identity with the adjacent first partial regions above, below, to the left or to the right of the first partial region, correcting the error using an asymmetrical characteristic of adjacent first partial regions in the horizontal direction.

28. The method of claim 27, wherein the predetermined angle is 90 degrees.

29. The method of claim 1, wherein estimating the representative orientations of the second partial regions comprises:

measuring curvatures of the first partial regions by measuring a vertical orientation difference with the representative orientation of adjacent first partial regions in the vertical direction, and a horizontal orientation difference with the representative orientation of adjacent first partial regions in the horizontal direction; and

estimating representative orientations of the second partial regions according to whether the first partial region has a large curvature or a small curvature.

30. The method of claim 29, wherein the vertical orientation difference is measured using the following equation:

Δ

v_θ

=∥

θ

<sub>[y−

1][x]</sub>−

θ

_[y+1][x]|−

90|

and the horizontal orientation difference is measured using the following equation;

Δ

h_θ

=∥

θ

<sub>[y][x−

1]</sub>−

θ

_[y][x+1]|−

90|

and the curvature is measured using the following equation;

Δ

Cv_θ

=α

Δ

v_θ

+β

Δ

h_θ

where Δ

vθ and

Δ

hθ

denote the vertical orientation difference and the horizontal orientation difference, respectively, Δ

Cvθ

denotes the curvature, and α and

β

denote proportionality constants determined according to characteristics of an image acquisition apparatus.

31. The method of claim 30, wherein it is determined that the curvature is small if the curvature is smaller than a predetermined curvature threshold, and it is determined that the curvature is large if the curvature is larger than the predetermined curvature threshold.

32. The method of claim 29, wherein estimating the representative orientation of the second partial regions comprises:

if the curvature of the first region to which the second partial regions belongs is determined to be small, allocating the representative orientations of the first partial regions as the representative orientations of the second partial regions; and

if the curvature of a first region to which a second partial regions belongs is determined to be large, selecting only gradients satisfying a predetermined pass condition among the gradients of pixels of the first partial region, and estimating the representative orientation of the second partial regions in response to the selected gradient.

33. The method of claim 32, wherein, in allocating the representative orientation of the first partial region as the representative orientation of the second partial regions, the representative orientation of the second partial regions is allocated in response to a vertical orientation difference with adjacent first partial regions in the vertical direction and a horizontal orientation difference with adjacent first partial regions in the horizontal direction.

34. The method of claim 1, further comprising, after estimating the representative orientation of the second partial regions:

determining whether or not an error exists in the estimated representative orientation of the second partial regions; and

if an error exists, dividing the second partial region for which the error exists into sub-partial regions of smaller sizes and then estimating representative orientations of the sub-partial regions.

35. The method of claim 34, further comprising:

dividing the sub-partial regions into sub-partial regions of smaller sizes until no error exists in the estimated representative orientation of the sub-partial regions.

36. The method of claim 34, wherein in determining whether or not an error exists, the curvature of the second partial regions is measured, and if the curvature is smaller than a predetermined curvature threshold, it is determined that an error exists.

37. An apparatus for estimating an orientation comprising:

a partial region division unit configured to divide an image into first partial regions;

an orientation estimation unit configured to estimate a representative orientation of the first partial regions by measuring gradients of pixels of the image, configured to filter the image using a double orientation filter according to the estimated representative orientation, and configured to reestimate the representative orientation of the first partial regions by remeasuring the gradients of pixels of the improved image; and

an orientation reestimation unit configured to divide the first partial regions into second partial regions, and configured to estimate the representative orientations of the second partial regions with respect to curvatures of the first partial regions in response to the remeasured representative orientations of the first partial regions and the remeasured gradients of the pixels.

38. The apparatus of claim 37, further comprising a uniformization unit configured to adjust a brightness of the image uniformly by processing brightness characteristics of the first partial regions, using a brightness characteristic of the image.

39. The apparatus of claim 38, wherein the image comprises a fingerprint pattern image, and wherein the uniformization unit comprises:

a fingerprint pattern uniformity degree determination unit configured to determine if a fingerprint pattern of the region is uniform, or else determining that the fingerprint pattern of the region is not uniform if a fingerprint pattern uniformity degree measured in response to an absolute value of a brightness difference among pixels in a first partial region is less than a predetermined uniformity degree threshold; and

a fingerprint pattern uniformization unit configured to low-pass filter a region determined to be not uniform.

40. The apparatus of claim 37, wherein the orientation estimation unit comprises:

a gradient measuring unit configured to perform convolution operations of predetermined neighboring pixels of a pixel whose gradient is being measured with gradient measuring operators of a first orientation and a second orientation in order to measure the first and second gradients; and

a representative orientation estimating unit configured to measure the orientation of pixels using the measured gradient of the pixels, and to estimate the representative orientations of the first partial regions using a histogram of the measured orientation for each pixel.

41. The apparatus of claim 40, wherein the representative orientation estimation unit comprises:

an orientation information measuring unit configured to measure the orientation of each pixel in response to a ratio of the first orientation gradient and the second orientation gradient;

a quantization unit configured to quantize pixels in predetermined angular units; and

a representative orientation determination unit configured to obtain the histogram of the quantized orientation and to determine an orientation corresponding to a location in which the histogram has a maximum value, as the representative orientation.

42. The apparatus of claim 40, wherein the orientation estimation unit further comprises:

a fingerprint structure improvement unit configured to determine whether or not the image is clear in the first partial regions and to filter images of the first partial regions using a double orientation filter according to whether or not the image is clear;

wherein the gradient measuring unit is further configured to remeasure the gradients of pixels of the improved image, and the representative orientation estimation unit is further configured to reestimate representative orientations of the first partial regions using the remeasured gradients.

43. The apparatus of claim 42, wherein the fingerprint structure improvement unit comprises:

a fingerprint structure clearness determination unit configured to measure a distinction coefficient and a single orientation coefficient in the first partial regions and configured to determine whether or not the image in the first partial regions is clear, by comparing the distinction coefficient and the single orientation coefficient with a predetermined distinction threshold and a single orientation threshold, respectively;

a first orientation filter configured to emphasize a boundary of ridges and valleys in the first partial regions in relation to four predetermined orientations;

a second orientation filter configured to reduce noise existing inside the ridges and valleys in relation to the four orientations; and

a image improvement unit configured to improve the image quality of a first partial region that is determined to be clear by filtering the first partial region that is determined to be clear with a first orientation filter and a second orientation filter for an orientation corresponding to the representative orientation of the first partial region, and that is configured to improve the image quality of a first partial region that is determined to be unclear by measuring clearness in the four orientations using the first orientation filter and the second orientation filter for each of the four orientations, and by filtering the first partial region that is determined to be unclear with the first orientation filter and the second orientation filter in an orientation in which the clearness is increased.

44. The apparatus of claim 37, further comprising a post-processing unit configured to measure an entire orientation of adjacent first partial regions and configured to correct an error of the remeasured representative orientations of the first partial regions in response to a difference of the entire orientations and the remeasured representative orientations of the first partial regions.

45. The apparatus of claim 44, wherein the post-processing unit comprises:

an adjacent region orientation evaluation unit configured to measure the entire orientations of the adjacent first partial regions in response to a mean value of a difference of the remeasured representative orientation and a predetermined angle;

an orientation difference measuring unit configured to measure a difference between the measured entire orientations and the remeasured representative orientations of the first partial regions;

an orientation error determination unit configured to determine whether or not an error occurs by comparing the difference between the measured entire orientation and the remeasured representative orientation of the first partial region, with a predetermined orientation difference threshold; and

an orientation error correction unit configured to correct the error in response to the mean value of adjacent first partial regions in the horizontal direction in relation to each first partial region in which an error occurs if there is a vertical orientation identity or a horizontal orientation identity with the adjacent first partial regions above, below, to the left or to the right of the first partial region, and if there is no vertical orientation identity or horizontal orientation identity with the adjacent first partial regions above, below, to the left or to the right of the first partial region in which an error occurs, configured to correct the error using an asymmetrical characteristic of the adjacent first partial regions in the horizontal direction.

46. The apparatus of claim 37, wherein the orientation reestimation unit comprises:

a partial region redivision unit configured to divide the first partial regions into second partial regions;

a representative orientation error measuring unit configured to measure a curvature of the first partial regions by measuring the vertical orientation difference with the representative orientation of the adjacent first partial regions in the vertical direction, and the horizontal orientation difference with the representative orientation of the adjacent first partial regions in the horizontal direction, and in response to the measured curvature, configured to measure the error of the representative orientation by determining whether the first partial region is a region of a large curvature or of a small curvature; and

a representative orientation reestimation unit configured to estimate representative orientations of the second partial regions according to whether the first partial region has a large curvature or a small curvature.

47. The apparatus of claim 46, wherein the representative orientation error estimation unit is further configured to determine whether or not an error exists in the estimated representative orientation of the second partial regions, and

the representative orientation error estimation unit is configured to output the estimated representative orientation as fingerprint information of the image if no error exists, and

the partial region redivision unit is configured to divide the second partial region into sub-partial regions of smaller sizes if an error exists, and the representative orientation reestimation unit is configured to estimate the representative orientation of the divided sub-partial regions according to whether or not an error exists.

48. The apparatus of claim 47, wherein the representative orientation error measuring unit is further configured to determine whether or not an error exists in the estimated representative orientation of each sub-partial region until no error exists in the estimated representative orientation of each sub-partial region,

and if no error exists in the representative orientation of each sub-partial region, the representative orientation error measuring unit is configured to output the representative orientation of each sub-partial region as the orientation information of the image, and

if an error exists in the representative orientation of the sub-partial regions, the partial region redivision unit is configured to divide the sub-partial regions into sub-partial regions of smaller sizes, and the representative orientation reestimation unit is configured to estimate the representative orientation of the divided sub-partial regions of smaller sizes.