Aligning rectilinear images in 3D through projective registration and calibration

• US 6,549,651 B2
• Filed: 02/14/2002
• Issued: 04/15/2003
• Est. Priority Date: 09/25/1998
• Status: Expired due to Term

• Current Assignee: Apple Computer Incorporated (Apple Inc.)
• Original Assignee: Apple Computer Incorporated (Apple Inc.)
• Inventors: Xiong, Yalin, Turkowski, Ken
• Primary Examiner(s): Mehta, Bhavesh
• Assistant Examiner(s): CHAWAN, SHEELA C
• Application Number:
  US10/077,102
• Publication Number:
  US 20020114536A1
• Time in Patent Office: 425 Days
• Field of Search: 382/305, 382/284, 382/294, 382/254, 382/154, 382/299, 382/260, 382/252, 382/253, 382/275, 382/276, 345/427, 345/433, 345/629
• US Class Current: 382/154
• CPC Class Codes: G06T 15/503 Blending, e.g. for anti-aliasing
• CPC Class Codes: G06T 15/506 Illumination models
• CPC Class Codes: G06T 3/4038 Image mosaicing, e.g. composing plane images from plane sub-images
• CPC Class Codes: G06T 7/32 using correlation-based methods
• CPC Class Codes: G06V 10/16 using multiple overlapping images; Image stitching
• CPC Class Codes: G06V 10/24 Aligning, centring, orientation detection or correction of the image

Abstract

An improved apparatus and method for creating high quality virtual reality panoramas is disclosed that yields dramatic improvements during the authoring and projecting cycles, with speeds up to several orders of magnitude faster than prior systems. In a preferred embodiment, a series of rectilinear images taken from a plurality of rows are pairwise registered with one another, and locally optimized using a pairwise objective function (local error function) that minimizes certain parameters in a projective transformation, using an improved iterative procedure. The local error function values for the pairwise registrations are then saved and used to construct a quadratic surface to approximate a global optimization function (global error function). The chain rule is used to avoid the direct evaluation of the global objective function, saving computation. In one embodiment concerning the blending aspect of the present invention, an improved procedure is described that relies on Laplacian and Gaussian pyramids, using a blend mask whose boundaries are determined by the grassfire transform. An improved iterative procedure is disclosed for the blending that also determines at what level of the pyramid to perform blending, and results in low frequency image components being blended over a wider region and high frequency components being blended over a narrower region. Human interaction and input is also provided to allow manual projective registration, initial calibration and feedback in the selection of photos and convergence of the system.

Claims

1. A method for creating a panoramic image from a plurality of rectilinear images obtained from a camera, the method comprising:

estimating a projective transformation during a pairwise registration process between at least two overlapping rectilinear images obtained from a camera, said projective transformation providing a Hessian matrix relating the at least two rectilinear images;

determining a vector of internal parameters and a vector of external parameters from the Hessian matrix to minimize image discrepancy in an overlapping region between the at least two overlapping rectilinear images during a process of global optimization; and

blending the at least two images to provide a smooth transition between the at least two images.

2. The method according to claim 1, wherein the projective transformation is a 3×

3 matrix comprising 9 projective parameters.

3. The method according to claim 1, wherein the pairwise registration process includes a local registration error function between the at least two images, $e_{\mathrm{ij}}=\frac{1}{A_{\mathrm{ij}}}$

∑

overlap

(sij

Ij

(Xj)+bij-Ii

(Mij

Xj))2.

4. The method according to claim 3, wherein the Hessian matrix, C_ij, between the at least two images is determined from the local registration error function,

5. The method according to claim 1, wherein the vector of internal parameters comprise of image center position, aspect ratio, and focal length.

6. The method according to claim 1, wherein the vector of external parameters represents the camera orientation with respect to a common frame of reference.

7. The method according to claim 4, wherein a global error function that is optimized during the global optimization process is defined by a function of the product of overlap area between the at least two images and the local registration error function.

8. The method according to claim 1, wherein the blending is performed such that high frequency signals in the overlapping region between the at least two images are blended over a narrow blend region range.

9. The method according to claim 1, wherein the blending is performed such that low frequency signals in the overlapping region between the at least two images are blended over a wide blend region range.

10. A system for creating a panoramic image from a plurality of rectilinear images obtained from a camera, the system comprising:

a processor for estimating a projective transformation during a pairwise registration process between at least two overlapping rectilinear images obtained from a camera, said projective transformation providing a Hessian matrix relating the at least two rectilinear images;

a global optimizer for determining a vector of internal parameters and a vector of external parameters from the Hessian matrix to minimize image discrepancy in an overlapping region between the at least two overlapping rectilinear images; and

a blender for blending the at least two images to provide a smooth transition between the at least two images.

11. The system according to claim 10, wherein the projective transformation is a 3×

3 matrix comprising 9 projective parameters.

12. The system according to claim 10, wherein the pairwise registration process includes a local registration error function between the at least two images, $e_{\mathrm{ij}}=\frac{1}{A_{\mathrm{ij}}}$

∑

overlap

(sij

Ij

(Xj)+bij-Ii

(Mij

Xj))2.

13. The system according to claim 12, wherein the Hessian matrix, C_ij, between the at least two images is determined from the local registration error function, $e_{\mathrm{ij}}$

(Mij)≈

eij0+(Mij-Mij0)T

Cij

(Mij-Mij0).

14. The system according to claim 10, wherein the vector of internal parameters comprise of image center position, aspect ratio, and focal length.

15. The system according to claim 10, wherein the vector of external parameters represents the camera orientation with respect to a common frame of reference.

16. The system according to claim 13, wherein a global error function that is optimized by the global optimizer is defined by a function of the product of overlap area between the at least two images and the local registration error function.

17. The system according to claim 10, wherein the blending is performed such that high frequency signals in the overlapping region between the at least two images are blended over a narrow blend region range.

18. The system according to claim 10, wherein the blending is performed such that low frequency signals in the overlapping region between the at least two images are blended over a wide blend region range.

19. A method for creating a panoramic image from a plurality of rectilinear images obtained from a camera, the method comprising:

estimating a projective transformation during a pairwise registration process between at least two overlapping rectilinear images obtained from a camera, said projective transformation providing a Hessian matrix relating the at least two rectilinear images;

determining a vector of internal parameters and a vector of external parameters from the Hessian matrix to minimize image discrepancy in an overlapping region between the at least two overlapping rectilinear images during a process of global optimization;

blending the at least two images to provide a smooth transition between the at least two images; and

mapping the blended images onto a three dimensional geometry surface.

20. The method according to claim 19, wherein the three dimensional geometry surface is a sphere.