V.R.S.Mani et al./ Elixir Adv. Engg. Info. 55A (2013) 13160-13163 13160
Introduction
The goal of image fusion is to integrate complementary
information from various modalities, so that the new image
should be more suitable for the purpose of human visual
perception and further analysis. Image fusion has become a
common term used within medical diagnostics and treatment.
The term fusion is used when multiple patient images are
registered and overlaid or merged to provide additional
information. Fused images may be created from multiple images
of the same imaging modality, or by combining information
from multiple modalities, such as Magnetic Resonanace Image
(MRI), Computed Tomography (CT), Positron Emission
Tomography (PET), and Single Photon Emission Computed
Tomography (SPECT). In radiology, these images serve
different medical purposes. For this reason, the radiologists
prefer integrating multiple imaging modalities to obtain more
details from input images. Commonly, a successful fusion
should extract complete information from the source images,
and should form the resulting image without introducing any
artifacts or inconsistencies.
Overview of Image Fusion
The medical image fusion mainly uses pixel based fusion
techniques. Usually, the pixel level fusion is broadly classified
into three main categories.
1. Spatial Domain Techniques: (PCA, Averaging, etc.)
2. Optimization Approach: (Bayesian Approach)
3. Transform Domain Approach: (Multi-resolution Techniques)
Initially, the fusion techniques are based on spatial domain
techniques. The basic techniques which were used are Principal
Component Analysis (PCA), Averaging, Weighted Averaging,
etc. the main advantage of this spatial domain technique is, they
are easy to implement. But there are few drawbacks in these
methods. This technique produces spatial distortion in images.
Also, some of the image details will not be present in the final
fused image with respect to the input images. The second one is
Bayesian approach, which suffers from the problem of
computational complexity. The third one is based on transform
domain approach .based on redundant multi resolution
decomposition techniques capable of preserving structural
characteristics. The transform decomposes the image into
several components and the various components are fused based
on their structural and functional importance and finally using an
inverse transform the fused image is reconstructed. It provides
more information for further analysis and diagnosis of various
diseases.. The different multi scale transform domain techniques
are wavelet, curvelet, contourlet, etc.
Proposed methods
There are some major drawbacks in the wavelet transform.
First, it doesn’t provide shift invariance, and it does not capture
the edges properly. Another major drawback in the wavelet
transform is, it provides limited information along the
horizontal, vertical and diagonal direction.
Dual Tree Complex Wavelet Transform
The above said drawbacks are removed using the proposed
technique. In the proposed technique Multimodal images are
decomposed using Dual Tree Complex Wavelet transform (DT-
CWT). 2D – DT-CWT is the combination of two 1-D
transforms. In wavelet transform, it has 1-D real filters. But in
dual tree, there are two trees containing complementary filter
values, one tree corresponds to real values and the other one is
imaginary. Designing complex filters is not an easy task, but it
makes the process more efficient. Also this Dual tree complex
wavelet transform produces Approximate Shift invariance and it
also provides limited directional information when compared to
the wavelet transform. Also this technique captures more edge
information when compared to wavelet transform.
Contourlet Transform
The Contourlet transform is used to decompose the image at
different scales and orientations. Contourlet transform is an
extension of wavelet transform and it uses directional filte
Tele:
E-mail addresses: vrsece@rediffmail.com
© 2013 Elixir All rights reserved
Multimodal image fusion using multiresolution techniques
V.R.S.Mani
1
, S. Arivazhagan
2
and J. Jason Braino
3
1
Department of ECE, NEC, Kovilpatti.
2
Department of ECE, Mepco Schlenk Engg. College, Sivakasi.
3
Communication Systems, NEC, Kovilpatti.
ABSTRACT
Multimodal Image Fusion techniques combine information from different sensors together
to produce a more accurate and efficient representation which is more useful for further
analysis. A Multi resolution based Multimodal Image Fusion Algorithm is this paper, an
automatic algorithm based on multi resolution technique for fusing multimodal images is
proposed. The multimodal images are decomposed using the Dual Tree Complex Wavelet
Transform and Contourlet Transform and they are fused using some efficient and robust
fusion rules. Finally, they are reconstructed using the Inverse transform and a new fused
image with more information content is obtained. The basic idea of all multiresolution
fusion schemes is motivated by the human visual system being primarily sensitive to local
contrast changes, e.g. the edges or corners. These techniques also provide better directional
information. The DT-CWT and Contourlet Transform methods are good at faithfully
retaining the salient structural information present in both the multimodal images.
© 2013 Elixir All rights reserved.
ARTICLE INFO
Article history:
Received: 7 June 2012;
Received in revised form:
13 February 2013;
Accepted: 22 February 2013;
Keywords
Image Fusion,
Dual Tree Complex Wavelet
Transform,
Contourlet Transform.
Elixir Adv. Engg. Info. 55A (2013) 13160-13163
Advanced Engineering Informatics
Available online at www.elixirpublishers.com (Elixir International Journal)