26-11-2012, 12:41 PM
A MULTI-MODAL AUTOMATIC IMAGE REGISTRATION TECHNIQUE BASED ON COMPLEX WAVELETS
A MULTI-MODAL AUTOMATIC.pdf (Size: 345.92 KB / Downloads: 49)
ABSTRACT
Image registration is considered one of the most fundamental and
crucial pre-processing tasks in digital imaging. This paper
describes a fast multimodal automatic image registration algorithm
that handles the alignment of IR and visible images. A multiresolution
approach based on Dual Tree-complex wavelet
transform is employed to speed up the process. At the coarsest
level, an accurate registration estimate for higher levels is
achieved, using edge detection and cross correlation. Mutual
Information, on the other hand, is applied at higher levels as a
matching criterion applied to the six orientation bands of the
complex wavelet. The process is completely automatic, and was
tested on several sets of synthetic and real data. Experimental
results show that the proposed technique exhibits better accuracy
than DWT-based algorithms for uni and multi-modal cases.
INTRODUCTION
Image Registration is the process of geometrically aligning
two or more images acquired from different viewpoints
(MultiView registration), at different times (Mutli-Temporal
registration), by different sensors (Multi-Modal registration), or a
combination of two or more of the aforementioned. In MultiView
analysis, the images may differ in translation, rotation, scaling, or
more complex transformations mainly due to camera positions,
while in multi-temporal analysis, images of the same scene may be
acquired at different times or under different lighting conditions,
and finally, in multimodal analysis, images are acquired by
different types of imagers or sensors.
A plethora of proposed algorithms can be found for image
registration which has gained, and is still receiving a lot of
attention in the research community, due to its importance and
necessity in many applications such as remote sensing, image
mosaicing, image fusion (Surveillance, historical monument
preservation), medicine (change detection, tumor growth
monitoring) and computer vision (Target tracking). A detailed
survey on conventional and newly proposed registration
algorithms can be found in [1].
RELATED WORK
Automatic registration has been extensively researched in the
past 20 years; however, this section covers the main proposed
schemes that employ multi-resolution processing, Mutual
Information, or the combination of both. The idea of addressing
the registration problem by applying coarse-to-fine resolution
strategy has proven to be an elegant method to speed up the whole
process while preserving, if not enhancing the accuracy of the
algorithm. In [3], a mutli-resolution scheme based on Discrete
Wavelet transform (DWT) is employed to register satellite images.
Maximum Modulus Maxima is applied on the LH and HL
frequency bands to extract edge points, and correlation is then
applied for matching. The authors in [4] developed a parallel
algorithm using the maxima of DWT coefficients for the feature
space, and correlation for the search space. Despite their achieved
performance, the above mentioned methods operate directly on
gray intensity values and hence they are not suited for handling
multi-sensor images. Mutual information methods on the other
hand, originating with Viola and Wells [5], are able to register
multimodal images since MI represents a measure of statistical
dependency between the reference and the senses images rather
than gray intensity values, which vary when different types of
imagers are used, or under different lighting conditions. A
multimodal brain image registration is developed and presented in
[6]. It combines the sum of difference (SAD) and the mutual
information (MI) into a matching criterion to enhance the
registration accuracy.
CONCLUSION
In this paper, a new technique for multimodal automatic
image registration algorithm is presented. To speed up the
processing, the algorithm is employed in a pyramidal fashion
based on Dual tree complex wavelet transform. At the lowest level,
edge maps are extracted and the matching is based on cross
correlation measure. The search interval is then refined for higher
levels employing Mutual Information as a matching criterion due
to its ability to register multimodal images and its light
computational load. The developed technique handles multi-modal
as well as uni-modal cases and has shown to have superior
accuracy when compared to its DWT counterpart.