Image Processing Reference
In-Depth Information
This work was supported by the Technology Innovation Program (or Technology Innovation
Program, “0043358, Information Composition and Recognition System for surrounding im-
ages possible for top view and panorama view of resolving power less than 10 cm) funded
By the Ministry of Trade, Industry & Energy (MI, Korea)” and “R&D Infrastructure for Green
Electric Vehicle (RE-EV) through the Ministry of Trade Industry & Energy (MOTIE) and Korea
Institute for Advancement of Technology (KIAT).”
1 Introduction
A driver assistance system to reduce the traffic accident rate analyzes data collected from vari-
ous sensors installed in a vehicle and provides its result to the driver. Around View Monitor
(AVM) system is used to prevent collision accidents during parking and driving by compos-
ing data received from more than three cameras installed around the vehicle and providing
surrounding image information to driver.
In general, a passenger car is equipped with four cameras with wide-angle lenses mounted
on the front, rear, left, and right sides of a vehicle to capture the maximum field of horizontal
and vertical images from the surroundings. A wide-angle lens, which can capture a picture
with a wide angle greater than 120° with a short focal length, generates radial distortion by
which a ray of light that enters the lens farther from its center is more curved than a ray of
light that enters the lens closer to its center due to the effect of a curved lens. This phenomenon
often occurs in fish-eye lenses used in AVM systems, and such distortion is more severe near
the edge of the image than at the center.
In order to correct the distortion of a fish-eye lens, two methods have been used: approxim-
ation of distortion functions into polynomial forms and the field of view (FOV) model, which
is a geographical model based on nonlinear distortion characteristics. In the polynomial dis-
tortion model, computational complexity increases as the order of polynomials increases and
the difficulty of application increases as the view angle of the fish-eye lens becomes larger.
However, the FOV model is more efficient than the polynomial distortion model because its
design is based on the nonlinear distortion of fish-eye lenses, although it might have addition-
al distortion when there is an error in the location of the distortion center.
To solve this problem, this chapter proposes a method to estimate a distortion center to
accurately correct the distortion that occurs in the camera image signals from an ultra-wide
viewing angle greater than 190°. In particular, this chapter proposes a method to find the cen-
ter point of distortion that can minimize distortion using a latice-paterned 2D plane and the
FOV distortion model to correct distortion.
The remainder of this chapter is organized as follows. In Section 2 , previous studies related
to distortion correction are described, whereas the proposed distortion center estimation
method using 2D planes is described in Section 3 . In Section 4 , the experiment environment
and results are discussed, and the conclusion is presented in Section 5 .
2 Related research
A complex calculation is required to transform 3D camera images into 2D planar images that
can be processed by computers. To reduce such computational complexity, a pinhole camera
model is used. A pinhole camera, which was used to locate the center point of an image, con-
verts 3D information into 2D planar pixel units based on the optical image received through
Search WWH ::

Custom Search