Fingerprint region segmenting apparatus, directional filter unit and methods thereof

• US 20060147096A1
• Filed: 01/04/2006
• Published: 07/06/2006
• Est. Priority Date: 01/05/2005
• Status: Abandoned Application

• Current Assignee: Samsung Electronics Co. Ltd.
• Original Assignee: Samsung Electronics Co. Ltd.
• Inventors: Lee, Dong-jae, Park, Deok-soo
• Application Number:
  US11/324,380
• Publication Number:
  US 20060147096A1
• Time in Patent Office: Days
• US Class Current: 382/124
• CPC Class Codes: G06V 40/1347 Preprocessing; Feature extraction

Abstract

A fingerprint region segmenting apparatus and methods thereof The fingerprint region segmenting apparatus may include at least one directional filter receiving an input fingerprint image and filtering the input fingerprint image to generate at least one directional image, a normalization unit normalizing the at least one directional image and a region classification unit dividing the normalized at least one directional image into a plurality of blocks and classifying each of the plurality of blocks. In an example, the classification for each of the plurality of blocks may be one of a foreground of the input fingerprint image and a background of the input fingerprint image. In an example method, a fingerprint may be segmented by segmenting a fingerprint image into a plurality of regions based on a plurality of directional images, each of the plurality of directional images associated with a different angular direction.

Claims

1. A fingerprint region segmenting apparatus, comprising:

a directional filter unit receiving an input fingerprint image and filtering the input fingerprint image to generate at least one directional image;

a normalization unit normalizing the at least one directional image; and

a region classification unit dividing the normalized at least one directional image into a plurality of blocks and classifying each of the plurality of blocks.

2. The fingerprint region segmenting apparatus of claim 1, further comprising:

a pre-processing unit for reducing noise in the input fingerprint image.

3. The fingerprint region segmenting apparatus of claim 1, wherein the one directional filter unit includes a plurality of directional filters and the at least one directional image includes a plurality of directional images.

4. The fingerprint region segmenting apparatus of claim 3, wherein the plurality of directional filters filters the input fingerprint image at a plurality of angular directions.

5. The fingerprint region segmenting apparatus of claim 4, wherein each of the plurality of directional filters filters the input fingerprint image at a different one of the plurality of angular directions.

6. The fingerprint region segmenting apparatus of claim 1, wherein the region classification unit classifies based at least in part on variances and symmetrical coefficients associated with the plurality of blocks.

7. The fingerprint region segmenting apparatus of claim 1, wherein the region classification unit classifies each of the plurality of blocks as being associated with one of a foreground of the input fingerprint image and a background of the input fingerprint image.

8. The fingerprint region segmenting apparatus of claim 4, wherein the plurality of angular directions includes at least one of 0°

, 45°

, 90°

, and 135°

.

9. The fingerprint region segmenting apparatus of claim 4, wherein the plurality of angular directions include a first angular direction, a second angular direction, a third angular direction and a fourth angular direction, wherein a brightness difference between pixels in the input fingerprint image for the first, second, third and fourth angular directions may be represented respectively as

$\begin{array}{c}{\mathrm{DGF}}_0\end{array}$

( x , y ) = ∑

k = - m m ⁢

{ I ⁡

( x + d , y - k ) - I ⁡

( x - d , y - k ) } DGF 45 ⁡

( x , y ) = ∑

k = - m m ⁢

{ I ⁡

( x + d 2 + k , y + d 2 - k ) - I ⁡

( x + d 2 + k , y - d 2 - k ) } DGF 90 ⁡

( x , y ) = ∑

k = - m m ⁢

{ I ⁡

( x - k , y + d ) - I ⁡

( x - k , y - d ) } DGF 135 ⁡

( x , y ) = ∑

k = - m m ⁢

{ I ⁡

( x - d 2 + k , y + d 2 - k ) - I ⁡

( x + d 2 + k , y - d 2 - k ) }

where DGF₀, DGF₄₅, DGF₉₀, and DGF₁₃₅ denote the brightness differences in angular directions of 0°

, 45°

, 90°

, and 135°

, respectively, coordinate (x,y) denotes coordinates indicating the position of the pixel in the directional image, and d denotes a distance from the pixel and (2m+1) denotes the width of a corresponding directional filter.

10. The fingerprint region segmenting apparatus of claim 9, wherein m equals 1 and d equals 2.

11. The fingerprint region segmenting apparatus of claim 1, wherein the normalization unit generates the normalized at least one directional image by normalizing brightness differences of each pixel of the at least one directional image into values in a given range.

12. The fingerprint region segmenting apparatus of claim 11, wherein the given range ranges from 0 to A, and the normalized brightness difference is expressed as

$\mathrm{NDGI}$

 

⁢

θ

⁡

( x , y ) = { min - DGF θ

⁡

( x , y ) min ×

( A + 1 ) 2 , if ⁢

 

⁢

DGF θ

⁡

( x , y ) <

0 max + DGF θ

⁡

( x , y ) max ×

( A + 1 ) 2 , otherwise

where NDGI denotes the normalized brightness difference, min denotes a brightness difference corresponding to a lowest 1% from among a brightness distribution, θ

denotes one of a plurality of angular directions associated with the at least one directional filter, and max denotes the brightness difference corresponding to a highest 1% among the brightness distribution.

13. The fingerprint region segmenting apparatus of claim 12, wherein A equals 255.

14. The fingerprint region segmenting apparatus of claim 1, wherein the region classification unit includes:

a block segmenting unit dividing the normalized directional image into the plurality of blocks, each of the plurality of blocks having a given size;

a variance calculation unit calculating a first variance of normalized brightness differences in each of the plurality of blocks;

a symmetrical coefficient calculation unit calculating a symmetrical coefficient of the normalized brightness difference in each of the plurality of blocks; and

a region determination unit determining a classification associated with each of the plurality of blocks based at least in part on the calculated first variance and the calculated symmetrical coefficient.

15. The fingerprint region segmenting apparatus of claim 14, wherein the variance calculation unit calculates a mean of the normalized brightness differences at a plurality of angular directions for each of the plurality of blocks, calculates a second variance of the normalized brightness differences at the plurality of angular directions for each of the plurality of blocks and selects a maximum value among the calculated second variances at the plurality of angular directions as the first variance for one of the plurality of blocks.

16. The fingerprint region segmenting apparatus of claim 15, wherein the mean is expressed as

$E_i$

( p , q ) = 1 mm ⁢

∑

x = pm + 1 pm + m ⁢

∑

y = qm + 1 qm + m ⁢

NDGI i ⁡

( x , y )

where coordinate (p,q) denotes a position of one of the plurality of blocks in the normalized at least one image and i denotes one of the plurality of angular directions.

17. The fingerprint region segmenting apparatus of claim 15, wherein the second variance is expressed as

$V_i$

( p , q ) = 1 mm ⁢

∑

x = pm + 1 pm + m ⁢

∑

y = qm + 1 qm + m ⁢

{ E i ⁡

( p , q ) - NDGI i ⁡

( x , y )

where coordinate (p,q) denotes a position of one of the plurality of blocks in the normalized at least one image and i denotes one of the plurality of angular directions.

18. The fingerprint region segmenting apparatus of claim 14, wherein the symmetrical coefficient calculation unit calculates the symmetrical coefficient for each of the plurality of blocks based on a ratio of a number of the normalized brightness differences greater than a central value in a brightness distribution to a number of the normalized brightness differences less than the central value in the brightness distribution.

19. The fingerprint region segmenting apparatus of claim 18, wherein the symmetrical coefficient is expressed as

$\mathrm{HS}$

( p , q ) = 

CHH ⁡

( p , q ) - CHL ⁡

( p , q ) 

CHH ⁡

( p , q ) + CHL ⁡

( p , q )

where coordinate (p,q) denotes a position of one of the plurality of blocks in the normalized at least one image, CHL denotes the number of normalized brightness differences less than the central value in the brightness distribution, and CHH denotes the number of normalized brightness differences greater than the central value in the brightness distribution.

20. The method of claim 14, wherein the region determination unit classifies a given block as associated with a foreground of the input fingerprint image if the variance is greater than a variance threshold and the symmetrical coefficient is less than a symmetrical coefficient threshold and classifies the given block as associated with a background of the input fingerprint image if the variance is not greater than a variance threshold and the symmetrical coefficient is not less than a symmetrical coefficient threshold.

21. The fingerprint region segmenting apparatus of claim 14, further comprising:

a preprocessing unit reducing noise in the input fingerprint image.

22. The fingerprint region segmenting apparatus of claim 21, wherein the preprocessing unit reduces the noise with a Gaussian-filtering process.

23. The fingerprint region segmenting apparatus of claim 1, further comprising:

a post-processing unit correcting a classification for at least one incorrectly classified block from among the plurality of blocks.

24. The fingerprint region segmenting apparatus of claim 23, wherein the at least one corrected block is initially classified incorrectly by the region classification unit.

25. The fingerprint region segmenting apparatus of claim 24, wherein the post-processing unit corrects the at least one incorrectly classified block by repeatedly median-filtering the fingerprint image in which the incorrectly classified block is classified.

26. A method of segmenting a fingerprint image, comprising:

filtering an input fingerprint image to generate at least one directional image;

normalizing the at least one directional image;

dividing the at least one normalized directional image into a plurality of blocks; and

classifying each of the plurality of blocks.

27. The method of claim 26, further comprising:

preprocessing the input fingerprint image to reduce noise before the filtering.

28. The method of claim 27, wherein the filtering filters the input fingerprint image at a plurality of angular directions.

29. The method of claim 27, wherein the dividing is based at least in part on a variance and a symmetrical coefficient of each of the plurality of blocks.

30. The method of claim 27, wherein the classifying classifies each of the plurality of blocks as being associated with one of a foreground of the input fingerprint image and a background of the input fingerprint image.

31. The method of claim 28, wherein the plurality of angular directions include at least one of 0°

, 45°

, 90°

, and 135°

.

32. The method of claim 28, wherein the plurality of angular directions include a first angular direction, a second angular direction, a third angular direction and a fourth angular direction, wherein a brightness difference between pixels in the input fingerprint image for the first, second, third and fourth angular directions may be represented respectively as

$\begin{array}{c}{\mathrm{DGF}}_0\end{array}$

( x , y ) = ∑

k = - m m ⁢

{ I ⁡

( x + d , y - k ) - I ⁡

( x - d , y - k ) } DGF 45 ⁡

( x , y ) = ∑

k = - m m ⁢

{ I ⁡

( x + d 2 + k , y + d 2 - k ) - I ⁡

( x + d 2 + k , y - d 2 - k ) } DGF 90 ⁡

( x , y ) = ∑

k = - m m ⁢

{ I ⁡

( x - k , y + d ) - I ⁡

( x - k , y - d ) } DGF 135 ⁡

( x , y ) = ∑

k = - m m ⁢

{ I ⁡

( x - d 2 + k , y + d 2 - k ) - I ⁡

( x + d 2 + k , y - d 2 - k ) }

where DGF₀, DGF₄₅, DGF₉₀, and DGF₁₃₅ denote the brightness differences in angular directions 0°

, 45°

, 90°

, and 135°

, respectively, coordinate (x,y) denotes coordinates indicating the position of the pixel in the directional image, and d denotes a distance from the pixel and (2m+1) denotes the width of a corresponding directional filter.

33. The method of claim 32, wherein m equals 1 and d equals 2.

34. The method of claim 26, wherein the normalizing includes normalizing brightness differences of each pixel of the at least one directional image into values in a given range.

35. The fingerprint region segmenting apparatus of claim 34, wherein the given range ranges from 0 to A, and the normalized brightness difference is expressed as

$\mathrm{NDGI}$

 

⁢

θ

⁡

( x , y ) = { min - DGF θ

⁡

( x , y ) min ×

( A + 1 ) 2 , ifDGF θ

⁡

( x , y ) <

0 max + DGF θ

⁡

( x , y ) max ×

( A + 1 ) 2 , otherwise

where NDGI denotes the normalized brightness difference, min denotes a brightness difference corresponding to a lowest 1% from among a brightness distribution, θ

denotes one of a plurality of angular directions associated with the at least one directional filter, and max denotes the brightness difference corresponding to a highest 1% from among the brightness distribution.

36. The method of claim 35, wherein A equals 255.

37. The method of claim 26, wherein the classifying includes:

dividing the at least one normalized directional image into the plurality of blocks, each of the plurality of blocks having a given size;

calculating a first variance of a normalized brightness differences for each of the plurality of blocks;

calculating a symmetrical coefficient of the brightness difference for each of the plurality of blocks; and

determining whether a classification associated with each of the plurality of blocks based on the calculated first variance and the calculated symmetrical coefficient.

38. The method of claim 37, wherein the calculating of the first variance includes:

calculating a mean of the normalized brightness differences at a plurality of angular directions for each of the plurality of blocks;

calculating a second variance of the normalized brightness differences at the plurality of angular directions for each of the plurality of blocks; and

selecting a maximum value among the calculated second variances at the plurality of angular directions as the first variance for one of the plurality of blocks.

39. The method of claim 37, wherein the mean is expressed as

$E_i$

( p , q ) = 1 mm ⁢

∑

x = pm + 1 pm + m ⁢

∑

y = qm + 1 qm + m ⁢

NDGI i ⁡

( x , y )

where coordinate (p,q) denotes a position of one of the plurality of blocks in the normalized at least one image and i denotes one of the plurality of angular directions.

40. The method of claim 37, wherein the second variance is expressed as

$V_i$

( p , q ) = 1 m ⁢

 

⁢

m ⁢

∑

x = pm + 1 pm + m ⁢

∑

y = qm + 1 qm + m ⁢

{ E i ⁡

( p , q ) - NDGI i ⁡

( x , y )

where coordinate (p,q) denotes a position of one of the plurality of blocks in the normalized at least one image and i denotes one of the plurality of angular directions.

41. The method of claim 37, wherein the calculating of the symmetrical coefficient includes is based on a ratio of a number of the normalized brightness differences greater than a central value to a number of the normalized brightness differences less than the central value.

42. The method of claim 37, wherein the symmetrical coefficient is expressed as

$\mathrm{HS}$

( p , q ) = 

CHH ⁡

( p , q ) - CHL ⁡

( p , q ) 

CHH ⁡

( p , q ) + CHL ⁡

( p , q )

where coordinate (p,q) denotes a position of one of the plurality of blocks in the normalized at least one image, CHL denotes the number of normalized brightness differences less than the central value and CHH denotes the number of normalized brightness differences greater than the central value.

43. The method of claim 37, wherein the classifying classified a given block as associated with a foreground of the input fingerprint image if the variance is greater than a variance threshold and the symmetrical coefficient is less than a symmetrical coefficient threshold.

44. The method of claim 37, wherein the classifying classified a given block as associated with a background of the input fingerprint image if the variance is not greater than a variance threshold and the symmetrical coefficient is not less than a symmetrical coefficient threshold.

45. The method of claim 27, wherein the preprocessing is performed before the at least one directional image is generated.

46. The method of claim 27, wherein the preprocessing includes a Gaussian-filtering process.

47. The method of claim 26, wherein the classifying classifies at least one of the plurality of blocks incorrectly.

48. The method of claim 47, further comprising:

correcting the classification of the at least one incorrectly classified block.

49. The method of claim 48, wherein the correcting includes applying a median-filtering process repeatedly.