Method for estimation of fundamental matrix in implementing a stereo vision

• US 7,251,356 B2
• Filed: 11/10/2003
• Issued: 07/31/2007
• Est. Priority Date: 04/22/2003
• Status: Expired due to Fees

• Current Assignee: Industry-University Cooperation Foundation Hanyang University
• Original Assignee: Industry-University Cooperation Foundation Hanyang University
• Inventors: Seo, Jung-Kak, Hong, Hyun-Ki, Jho, Cheung-Woon
• Primary Examiner(s): Bali; Vikkram
• Application Number:
  US10/703,425
• Publication Number:
  US 20040213452A1
• Time in Patent Office: 1,359 Days
• Field of Search: 382/154, 382/103, 382/199, 345/419, 345/420
• US Class Current: 382/154
• CPC Class Codes: G06T 7/593 from stereo images

Abstract

A method for estimation of a fundamental matrix by selecting sets of correspondence points is provided. According to the method, an entire image is divided into several sub-regions, and the number of the inliers in each sub-region and the area of each region is examined. The standard deviation are used as quantitative measures to select a proper inlier set. This method achieves a more precise estimation of the fundamental matrix than conventional method does.

Claims

1. A method for estimation of a fundamental matrix by selecting a set of correspondence points in the camera images to obtain the relative geometrical information for the two cameras in implementing a stereo vision, the method comprising the steps of:

(a) selecting a predetermined number of random correspondence points, and then obtaining an initial fundamental matrix by linear method;

(b) calculating an error of each correspondence point from the initial fundamental matrix, and obtaining the median value of the set of the calculated errors;

(c) iterating for predetermined times the steps of (a) and (b), and storing the initial fundamental matrix and the median value according to the initial fundamental matrix for each iteration;

(d) selecting the least median value and the larger median value ranged within about 10% of the least median value among the stored median value, and corresponding fundamental matrix;

(e) presetting a threshold value corresponding to the respective selected median value, and obtaining inlier sets corresponding to the respective median value using the corresponding threshold value and the initial fundamental matrix;

(f) selecting the inlier set wherein the standard deviation of the distribution of the inlier set is minimum; and

(g) obtaining the fundamental matrix using the selected of inlier set.

2. The method according to claim 1, wherein the relation of the correspondence point x and x′

in the respective camera images, and the fundamental matrix F is as Equation 1;

x′

^TFx=0 

 

[Equation 1].

3. The method according to claim 1, wherein the number of random correspondence points to be selected for obtaining an initial fundamental matrix in step (a) is 7 or 8.

4. The method according to claim 1, wherein the error r of correspondence point in step (b) is calculated according to the Equation 2, and the median value means the value located in the middle among the errors arranged in the order of magnitude:

r=d(x, F^Tx′

)+d(x′

, Fx) 

 

[Equation 2]

where d(x, l) is the distance between a point x and a line l.

5. The method according to claim 1, wherein The number of iteration of steps (a) and (b) in the step (c) is determined by the following Equation 3:

N=log(1−

P)/log(1−

(1−

ε

)^q) 

 

[Equation 3]

where,

ε

is the ratio of the outlier to all points;

q is the number of the sample points; and

P is the probability that these points are the inliers in sampling q points at N times.

6. The method according to claim 1, wherein the threshold value in the step (e) is obtained according to the Equation 4:

τ

=2.5×

1.4826[1+5/(p−

q)]√

{square root over (Median)} 

 

[Equation 4]

where,

p is the number of correspondence points,

q is the number of the sample points; and

Median is the median value.

7. The method according to claim 1, wherein, the image is divided into sub-regions in step (f), and the standard deviation means the distribution degree regading point density or area density of correspondence points in the sub-regions.

8. The method according to claim 7, wherein the image divided into sub-regions by Equation 5, and the standard deviation of point density that represents a degree of the point distribution in each sub-region relative to the entire image is calculated by Equation 6:

W_s=W/int(√

{square root over (N))}

H_s=H/int(√

{square root over (N))} 

 

[Equation 5]

9. The method according to claim 7, wherein the image is divided into sub-regions in the shape of triangle using the Delaunay triangulation, and the standard deviation of the area density is calculated as Equation 8:

10. The method according to claim 1, wherein the fundamental matrix is obtained using iterative method from the inlier set selected in the step (g).