IMAGE PROCESSING APPARATUS AND METHOD

• US 20130011077A1
• Filed: 07/13/2012
• Published: 01/10/2013
• Est. Priority Date: 01/13/2010
• Status: Active Grant

• Current Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Original Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Inventors: MATSUMOTO, Nobuyuki, KAWATA, Satoshi, ISOGAWA, Kenzo, KANEKO, Toshimitsu
• Granted Patent: US 8,693,804 B2
• Time in Patent Office: Days
• US Class Current: 382/275
• CPC Class Codes: G06T 5/70 Denoising; Smoothing

Abstract

According to one embodiment, an image processing apparatus includes first and second computation portions, a selection portion, a projection portion, and a weighted averaging portion. The first computation portion is configured to obtain magnitudes of correlations between a first vector and plural basis vectors. The selection portion is configured to select basis vectors from the plural basis vectors. The projection portion is configured to select a second region, obtain a first projection vector by projecting the first vector onto a subspace formed by the selected basis vectors and obtain a second projection vector for each second region by projecting a second vector onto the subspace. The second computation portion is configured to compute a distance between the first and second projection vectors. The weighted averaging portion is configured to weighted average a pixel value of the second pixel to obtain an output pixel value of a first pixel.

Claims

1. An image processing apparatus comprising:

a first computation portion configured to obtain magnitudes of correlations between a first vector having, as elements, pixel values of pixels in a first region containing a first pixel in an image and a plurality of basis vectors;

a selection portion configured to select basis vectors from the plurality of basis vectors in accordance with the magnitudes of the correlations;

a projection portion configured to select second regions containing second pixels in the image, to obtain a first projection vector by projecting the first vector onto a subspace formed by the selected basis vectors and to obtain a second projection vector for each second region by projecting onto the subspace a second vector in which pixel values of pixels in each second region are arranged;

a second computation portion configured to compute, for each second region, a distance between the first projection vector and the second projection vector corresponding to each second region; and

a weighted averaging portion configured to weighted average a pixel value of the second pixel with giving larger weights to the second pixel as the distance corresponding to each second region is smaller so as to obtain an output pixel value of the first pixel.

2. The image processing apparatus according to claim 1, wherein the projection portion selects a plurality of second regions each containing the second pixel.

3. The image processing apparatus according to claim 1, wherein the first computation portion computes the magnitudes of the correlations based on inner products of the first vector and the plurality of basis vectors.

4. The image processing apparatus according to claim 2, wherein

the first computation portion computes the magnitudes of the correlations based on squared inner products of the first vector and the plurality of basis vectors, and

the selection portion adds the magnitudes of the correlations from the basis vector having the largest magnitude of the correlation until a sum of the added magnitudes of the correlations reaches a predetermined threshold value and selects the basis vectors having the magnitudes of the correlations which have been added until the sum of the added magnitudes of the correlations reaches the threshold value.

5. The image processing apparatus according to claim 2, wherein

the first computation portion computes the magnitudes of the correlations based on absolute values of inner products of the first vector and the plurality of basis vectors, and

the selection portion adds the magnitudes of the correlations from the basis vector having the largest magnitude of the correlation until a sum of the added magnitudes of the correlations reaches a predetermined threshold value and selects the basis vectors having the magnitudes of the correlations which have been added until the sum of the added magnitudes of the correlations reaches the threshold value.

6. The image processing apparatus according to claim 2, wherein

the selection portion selects the basis vectors in accordance with the magnitudes of the correlations from the plurality of basis vectors except basis vectors in each of which an absolute value of a sum of elements is larger than a predetermined criterion, and

the projection portion forms the subspace by the basis vectors in each of which the absolute value of the sum of the elements is larger than the predetermined criterion, and the selected basis vectors.

7. The image processing apparatus according to claim 1, further comprising:

a noise amount computation portion that computes an amount of noise in the first region containing the first pixel, wherein

the selection portion selects the basis vectors so that a number of the selected basis vectors is smaller as the amount of noise is larger.

8. The image processing apparatus according to claim 2, wherein the first computation portion computes the basis vectors by performing a principal component analysis on a plurality of blocks which are extracted from at least one image in advance.

9. The image processing apparatus according to claim 2, wherein the first computation portion uses base of a two-dimensional discrete wavelet transform as the basis vectors.

10. The image processing apparatus according to claim 2, wherein the first computation portion computes the basis vectors by performing an independent component analysis on a plurality of blocks which are extracted from at least one image in advance.

11. An image processing method comprising:

obtaining magnitudes of correlations between a first vector having, as elements, pixel values of pixels in a first region containing a first pixel in an image and a plurality of basis vectors;

selecting basis vectors from the plurality of basis vectors in accordance with the magnitudes of the correlations;

selecting second regions containing second pixels in the image;

obtaining a first projection vector by projecting the first vector onto a subspace formed by the selected basis vectors;

obtaining a second projection vector for each second region by projecting onto the subspace a second vector in which pixel values of pixels in each second region are arranged;

computing, for each second region, a distance between the first projection vector and the second projection vector corresponding to each second region; and

weighted averaging a pixel value of the second pixel with giving a larger weight to the second pixel as the distance corresponding to each second region is smaller so as to obtain an output pixel value of the first pixel.

12. The image processing method according to claim 11, wherein the selecting of the second regions selects a plurality of second regions each containing the second pixel.

13. The image processing method according to claim 11, wherein the obtaining of the magnitudes of the correlations includes computing the magnitudes of the correlations based on inner products of the first vector and the plurality of basis vectors.

14. The image processing method according to claim 12, wherein

the obtaining of the magnitudes of the correlations includes the magnitudes of the correlations based on squared inner products of the first vector and the plurality of basis vectors, and

the selecting of the basis vectors includes

adding the magnitudes of the correlations from the basis vector having the largest magnitude of the correlation until a sum of the added magnitudes of the correlations reaches a predetermined threshold value, and

selecting the basis vectors having the magnitudes of the correlations which have been added until the sum of the added magnitudes of the correlations reaches the threshold value.

15. The image processing method according to claim 12, wherein

the obtaining of the magnitudes of correlations includes computing the magnitudes of the correlations based on absolute values of inner products of the first vector and the plurality of basis vectors, and

the selecting of the basis vectors includes

adding the magnitudes of the correlations from the basis vector having the largest magnitude of the correlation until a sum of the added magnitudes of the correlations reaches a predetermined threshold value, and

selecting the basis vectors having the magnitudes of the correlations which have been added until the sum of the added magnitudes of the correlations reaches the threshold value.

16. The image processing method according to claim 12, wherein

the selecting of the basis vector includes selecting the basis vectors in accordance with the magnitudes of the correlations from the plurality of basis vectors except basis vectors in each of which an absolute value of a sum of elements is larger than a predetermined criterion, and

the obtaining of the first projection vector includes forming the subspace by the basis vectors in each of which the absolute value of the sum of the elements is larger than the predetermined criterion, and the selected basis vectors.

17. The image processing method according to claim 11, further comprising:

computing an amount of noise in the first region containing the first pixel, wherein

the selecting of the basis vector includes selecting the basis vectors so that a number of the selected basis vectors is smaller as the amount of noise is larger.

18. The image processing method according to claim 12, wherein the obtaining of the magnitudes of the correlations includes computing the basis vectors by performing a principal component analysis on a plurality of blocks which are extracted from at least one image in advance.

19. The image processing method according to claim 12, wherein the obtaining of the magnitudes of the correlations includes using base of a two-dimensional discrete wavelet transform as the basis vectors.

20. The image processing apparatus according to claim 12, wherein the obtaining of the magnitudes of the correlations includes computing the basis vectors by performing an independent component analysis on a plurality of blocks which are extracted from at least one image in advance.