THE RESEARCH AND IMPLEMENTATION
OF DYNAMIC EPIPOLAR REARRANGEMENT
BingXuan GUO a , YuanZheng SHAO\ ZhiChao ZHANG 8 , LingYan Dong 3
a State Key Lab of Information Engineering in Surveying Mapping and Remote Sensing,
Wuhan University, China - mobilemap@gmail.com
Commission WGS-PS, WG IV/9
KEY WORDS: Epipolar Image, Dynamic Epipolar Rearrangement, Image Pyramid, Stereoscopic Observation, Memory Pool
ABSTRACT:
Image re-sampling according to epipolar geometry is a prerequisite for a variety of photogrammetric tasks such as stereoscopic
observation. Executing stereoscopic observation by virtue of a pair of epipolar images is a common approach. This paper proposes a
new approach to generate epipolar images and measure in stereoscopic environment. In order to improve the accuracy of
stereoscopic observation, a dynamic epipolar rearrangement (DER) method is presented. In this method, there is no need to construct
the epipolar image pyramid, because the needed epipolar images for measurement is dynamically generated from the original stereo
pair pyramid. Then scale the epipolar images and transfer it to the screen buffer directly for stereoscopic observation. The key
technology and method for stereoscopic observation, such as constructing pyramid image data structure with high fidelity,
dispatching image data by using multi-threaded and memory pool strategy to accelerate the speed of image display, especially the
integration operation of integrating the generation of epipolar images with the transfer of the epipolar images to the screen buffer, are
deeply investigated.
1. INTRODUCTION
At present, digital image acquisition of high resolution,
accuracy and multi-spectrum and real-time imagery is coming
true, how quickly and efficiently to survey the interested targets
and obtain the information from these images has become an
active research topic in digital photogrammetry.
The traditional stereoscopic observation strategy is to generate
epipolar images after obtaining the stereo pair [1], then the
stereopair is tiled and the epipolar images pyramid data
structure are constructed from the original stereopair, which
create different layers with different resolution [6], [9]. In the
course of stereoscopic observation, the images in the epipolar
pyramid, which belong to different tiles and layers, are
scheduled with the change of the viewpoint.
There are three main truncation errors about the traditional
strategy. The first exists in the course of generating epipolar
images from the stereopair. The second error is introduced
when constructing epipolar images pyramid, which is because
the common pyramid algorithm have the loss of accuracy. The
third is due to pixel truncation when transfer the epipolar image
to screen buffer. The cumulative error caused by the above
three errors affect the accuracy that match the cursor with the
interested objects in stereoscopic observation, leading to the
decrease of measurement accuracy consequently.
To deal with these errors and restrictions, a cylindrical
rectification technique was proposed in [2] which used a
separate transformation for each epipolar line. However, the
technique was complex, omitted many implementation details
and worked largely in 3D. A later work, [3] overcame most of
these problems, by using the tools of oriented projective
geometry to perform a similar nonlinear rectification without
using 3D.
In this article, the key technology and method for stereoscopic
observation, such as pyramid image, multi-threaded and
memory pool strategy, will be researched. In order to improve
the accuracy of stereoscopic observation, a dynamic epipolar
rearrangement!DER) method is presented. In this method, there
is not necessary to construct the epipolar image pyramid,
because the needed epipolar images for measurement is
dynamically generated from the original stereopair pyramid.
Then scale the epipolar images and transfer it to the screen
buffer directly for stereoscopic observation.
In Sections 2 the related technologies, such as constructing
pyramid image data structure with high fidelity, dispatching
image data by using multi-threaded and memory pool, are
considered. Section 3 describes the strategy of dynamic
epipolar rearrangement. Section 4 presents the systematic
integration which illustrates the full process. A proving
experiment is also performed in section 5 and the conclusion is
drawn in section 6.
2. BACKGROUND
2.1 Epipolar geometry
The epipolar geometry is the intrinsic projective geometry
between two views. It is independent of scene structure, and
only depends on the cameras' internal parameters and relative
pose [7].
An epipolar line is the intersection of an epipolar plane with the
image plane. All epipolar lines intersect at the epipole. An
epipolar plane intersects the left and right image planes in
epipolar lines, and defines the correspondence between the lines.
The epipolar geometry between two views is essentially the
geometry of the intersection of the image planes with the pencil