SYSTEMS AND METHODS FOR REGISTRATION OF IMAGES

• US 20160350893A1
• Filed: 03/22/2016
• Published: 12/01/2016
• Est. Priority Date: 05/29/2015
• Status: Active Grant

• Current Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Original Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Inventors: FLETCHER, Peter Alleine
• Granted Patent: US 10,089,713 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06T 3/153 using elastic snapping

Abstract

A method for registering images aligns a fixed image with a corresponding moving image. A target image and a reference image are selected. The reference and target images are relatively low-res representations of the fixed and moving images respectively. A current image distortion map is determined that defines distortions between the target and reference images. A residual image is determined by comparing the reference image and a warped target image derived from the target image based on the current image distortion map. A residual image distortion map is determined based on transform coefficients of cosine functions fitted to the residual image. The coefficients are determined by applying a DCT to a signal formed by the residual image and image gradients of the warped target or reference image. The current image distortion map is combined with the residual image distortion map to align the fixed image and the moving image.

Claims

1. A method for aligning a fixed image and a corresponding moving image, the method comprising:

selecting a target image and a reference image, the target image being a relatively low resolution representation of the moving image and the reference image being a relatively low resolution representation of the fixed image;

determining a current image distortion map defining distortions between the selected target image and reference image;

determining a residual image by comparing the reference image and a warped target image derived from the target image based on the current image distortion map;

determining a residual image distortion map based on transform coefficients of cosine functions fitted to the residual image, the transform coefficients being determined by applying a discrete cosine transform to a signal formed by the residual image and image gradients of the warped target image or the reference image; and

combining the current image distortion map with the residual image distortion map to align the fixed image and the moving image.

2. The method of claim 1, further comprising:

determining a degree of distortion associated with at least one of the fixed image and the moving image using the current image distortion map; and

filtering the transform coefficients based on the degree of distortion.

3. The method of claim 1, wherein the transform coefficients are determined by calculating a discrete cosine transformation of at least one previously determined image gradient.

4. The method of claim 1, further comprising filtering the transform coefficients.

5. The method of claim 4, wherein the filter has a Gaussian function.

6. The method of claim 5, further comprising determining an improved distortion map by combining the current image distortion map and the residual image distortion map weighted by a current convergence factor, the current convergence factor being determined on a basis of at least two previous sets of transform coefficients.

7. The method of claim 1, wherein the residual image distortion map is produced using inverse discrete cosine transform applied to the transform coefficients.

8. The method of claim 1, wherein determining the residual image distortion map further comprises:

determining a convergence ratio for the current transform coefficient and at least one previous transform coefficient; and

determining a further convergence factor using a previous convergence factor based on said convergence ratio.

9. A method for transforming a target image with respect to a reference image, the method comprising:

determining a current image distortion map defining distortions between the target image and the reference image;

determining a residual image by comparing the reference image and a warped target image derived from the target image on a basis of the current image distortion map;

determining a residual image distortion map using a plurality of basis functions having transform coefficients determined by applying a discrete frequency transform to a signal formed based on the residual image and image gradients of the warped target image or the reference image; and

combining the current image distortion map and the residual image distortion map to form an improved distortion map, wherein the improved distortion map is used to transform the target image for displaying on a display device.

10. The method of claim 9, wherein the improved distortion map is determined by combining the current image distortion map and the residual image distortion map weighted by a current convergence factor, the current convergence factor being determined on a basis of at least two previous sets of transform coefficients.

11. The method of claim 9, further comprising:

determining a degree of distortion associated with at least one of the fixed image and the moving image using the current image distortion map; and

filtering the transform coefficients based on the degree of distortion.

12. A method for determining an image distortion map relating a fixed image and a corresponding moving image, the method comprising:

(a) selecting a target image and a reference image, the target image being a relatively low resolution representation of the moving image and the reference image being a relatively low resolution representation of the fixed image;

(b) determining an initial distortion map defining distortions between the target image and the corresponding reference image;

(c) using the initial distortion map to derive an initial warped target image from the target image;

(d) determining current transform coefficients by calculating a discrete frequency transform for a residual image and image gradients in the initial warped target image, wherein the residual image is formed by comparing the reference image and the initial warped target image;

(e) filtering the determined current transform coefficients to form a current residual image distortion map;

(f) determining an improved image distortion map using the initial image distortion map and the current residual image distortion map weighted by a current convergence factor, wherein the current convergence factor is determined on a basis of at least two previous sets of transform coefficients;

(g) applying the determined further image distortion map to the target image to derive a further warped target image; and

(h) repeating steps (d)-(g) for a predetermined number of iterations to determine the image distortion map.

13. The method of claim 12, further comprises:

(i) receiving a further target image and a further reference image having resolution higher than resolution of the initial target and reference images; and

(j) repeating steps (d)-(h) for the further target and further reference images using an image distortion map from a previous iteration as an initial image distortion map.

14. The method of claim 13, further comprising determining a further convergence factor using a convergence ratio and the current convergence factor.

15. The method of claim 12, wherein the initial distortion map is constructed from a rigid transformation of the selected target image and reference image.

16. A method for aligning a reference image and a target image, the method comprising:

receiving a first value of a numerical representation of a first plurality of motion vectors between the first image and the second image;

receiving a subsequent value of the numerical representation of a subsequent plurality of motion vector between the first image and the second image;

calculating a ratio of a difference between the first and subsequent value and a sum of the first and subsequent value; and

comparing the ratio with a threshold value to determine whether the process of iteratively aligning the target image to the reference image is converging or diverging, wherein the subsequent plurality of motion vectors are adjusted in response to the determination to align the target image to the reference image.

17. A method for aligning a reference image and target image, the method comprising:

iteratively aligning the target image to the reference image to determine a first plurality of motion vectors and a second plurality of motion vectors, wherein the second plurality of motion vectors is determined in a subsequent iteration of aligning the target image to the reference image;

combining motion vectors of the first plurality of motion vectors in accordance with a predetermined pattern to generate a first iteration characteristic value associated with a first iteration of aligning the target image and the reference image;

combining motion vectors of the second plurality of motion vectors in accordance with a predetermined pattern to generate a second iteration characteristic value associated with the subsequent iteration of aligning the target image and the reference image;

calculating a ratio of a difference between the first and second iteration characteristic values and a sum of the first and second iteration characteristic values; and

comparing the ratio with a threshold value to determine whether the process of iteratively aligning the target image to the reference image is converging or diverging, wherein the second plurality of motion vectors are adjusted in response to the determination to align the target image to the reference image.

18. The method according to claim 17, wherein motion vectors in the second plurality of motion vectors are amplified linearly if the process of iteratively aligning the images is converging and attenuated exponentially otherwise.