Correcting distortion in an imaging system using parametric motion estimation

• US 6,597,816 B1
• Filed: 10/30/1998
• Issued: 07/22/2003
• Est. Priority Date: 10/30/1998
• Status: Expired due to Fees

• Current Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Original Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Inventors: Altunbasak, Yucel, Patti, Andrew
• Primary Examiner(s): ROGERS, SCOTT A
• Application Number:
  US09/183,820
• Time in Patent Office: 1,726 Days
• Field of Search: 382/275, 382/107, 382/254, 382/274, 382/291, 382/286, 348/208.1, 348/241, 348/352, 348/607, 348/620, 348/335
• US Class Current: 382/275
• CPC Class Codes: G06T 5/80 Geometric correction
• CPC Class Codes: H04N 25/61 the noise originating only from the lens unit, e.g. flare, shading, vignetting or "cos4"
• CPC Class Codes: H04N 5/145 Movement estimation for video coding H04N19/51

Abstract

Distortion correction in an imaging system using parametric motion estimation. Parametric motion estimation is used to correct for geometric lens distortion and illumination changes. Models for geometric lens distortion and different illumination models are presented, as is a method for an iterative linearising method of solving the resulting sets of nonlinear equations.

Claims

1. A method for correcting distortion in an imaging system, the method comprising:

a) applying a motion estimation model to a first and a second image to generate a set of nonlinear equations representing distortion, wherein the motion estimation model accommodates a change in illumination along at least one motion trajectory between the first and second images;

b) solving the set of nonlinear equations to produce a set of correction parameters; and

c) correcting at least one of the first and second images using the correction parameters to remove distortion.

2. The method of claim 1 where the distortion includes geometric lens distortion.

3. The method of claim 2 wherein geometric lens distortion is modeled by the transform:

$x_g=\frac{x}{1+{\mathrm{\kappa <br><br>}}_7<br><br>\left(x^2+y^2\right)}<br><br>\text{\hspace{1em}<br><br>}<br><br>y_g=\frac{y}{1+{\mathrm{\kappa <br><br>}}_7<br><br>\left(x^2+y^2\right)}$

where (x,y) are the corrected pixel locations, (x_g,y_g) are distorted pixel locations, and κ

₇ is a geometric lens distortion factor.

4. The method of claim 1 wherein the correction parameters correct a change in illumination along the at least one motion trajectory.

5. The method of claim 4 wherein illumination changes are modeled with a linear model.

6. The method of claim 4 wherein illumination changes are modeled with a Vignetting model.

7. The method of claim 1 where the method of solving the set of nonlinear equations is a multivariable nonlinear optimization method.

8. The method of claim 1 wherein step b) further comprises:

a) selecting an initial set of solution parameters;

b) linearizing the set of nonlinear equations around the set of solution parameters;

c) solving the resulting linearized set of equations to reach an updated set of solution parameters; and

d) repeating steps b) and c) until a condition defined by a selected one of a finite number of iterations and a predetermined convergence criteria is met.

9. The method of claim 8 where the set of nonlinear equations includes correction for geometric lens distortion.

10. The method of claim 8 where the set of nonlinear equations includes correction for illumination changes.

11. The method of claim 8 where the set of nonlinear equations includes correction for both geometric lens distortion and illumination changes.

12. The method of claim 11 where the model for illumination changes is a linear model.

13. The method of claim 11 where the model for illumination changes is a Vignetting model.

14. The method of claim 11 wherein geometric lens distortion is modeled by the transform:

$x_g=\frac{x}{1+{\mathrm{\kappa <br><br>}}_7<br><br>\left(x^2+y^2\right)}<br><br>\text{\hspace{1em}<br><br>}<br><br>y_g=\frac{y}{1+{\mathrm{\kappa <br><br>}}_7<br><br>\left(x^2+y^2\right)}$

where (x,y) are the corrected pixel locations, (x_g,y_g) are distorted pixel locations, and κ

₇ is the distortion factor.

15. A computer storage medium having instructions stored therein for programming a computer to perform the method of claim 11.

16. The method of claim 1 wherein step a) further comprises the step of performing motion estimation modeling using at least one of a translational motion model, an affine motion model, a bilinear motion model, and a perspective motion model.