29-06-2013, 12:16 PM
A presentation onImage Registration for Perspective Transformation
Image Registration.ppt (Size: 4.47 MB / Downloads: 24)
What is Image Registration ?
It is a process of aligning two or more images.
It involves locating and matching similar region in two or more images to be registered.
Classification-:
Multimodal-(from different sensors)
Multi Temporal-(from different time)
Viewpoint registration
Template registration
Steps in Image Registration
Preprocessing
Feature Selection
Feature Correspondence
Determination of a Transformation Function
Re sampling
Applications
Remote sensing-:agriculture and crop forecasting, water urban planning, range land monitoring, minerals and oil exploration, cartography, disease control.
Medical-:tumor monitoring, detection of anatomical changes.
Industrial image analysis systems
Projective Geometry
When we consider the imaging process of a camera, it becomes clear that Euclidean geometry is insufficient.
Lengths and angles are no longer preserved, and parallel lines may intersect.
Euclidean geometry is actually a subset of what is known as projective geometry. In fact, there are two geometries between them: similarity and affine.
Projective ⊃ affine⊃ similarity⊃ Euclidian
Vanishing line & point
Because image formation is the projection from a 3D world to a 2D surface, each point on the image plane is the projection of an infinite number of points in the world
In perspective drawing, lines on the paper which represent parallel lines in the world intersect on the paper at a point known as the vanishing point
Objective
(1) For each point in the first image determine the
corresponding point in the second image
(this is a search problem)
(2) For each pair of matched points determine the 3D
point by triangulation
(this is an estimation problem)
Basic concept
This algorithm is to take advantage of a geometric depth constraint, which is stronger than the epipolar constraint, to limit a feature correspondence point within a very small area along a horizontal epipolar line. And then the normalized cross correlation is performed in the area to search a correspondence point.
Analysis & Conclusion
Using the epipolar geometry and rectification concepts image registration become easy as the matching the feature points in two images and computational requirement is less.
We can also increase some more effective correlation algorithms for searching the match point in the target image like “Fast Normalized Cross-correlation” and “wavelets convolution” so as to further improve the match rate and match speed