29-06-2012, 06:22 PM
Multi-scale Image Fusion Using the Parameterized Logarithmic Image Processing Model
Multi-scale Image Fusion.pdf (Size: 1.83 MB / Downloads: 29)
Abstract
Image fusion is the process of combining multiple
images into a single image which retains the most pertinent
information from each original image source. More recently,
multi-scale image fusion approaches have emerged as a means of
providing a more meaningful fusion which better reflects the
human visual system. In this paper, multi-scale decomposition
techniques and image fusion algorithms are adapted using the
Parameterized Logarithmic Image Processing (PLIP) model, a
nonlinear image processing framework which more accurately
processes images. Experimental results via computer simulations
illustrate the improved performance of the proposed algorithms
by both qualitative and quantitative means.
INTRODUCTION
Image fusion algorithms fuse multiple source images of the
same scene into a single image. The original source images
may vary in sensor type, resolution, focal point, etc., and
therefore can vary greatly in appearance from one another [1].
J\dditionally, it should be noted that the source images to be
fused are assumed to be registered. Image fusion algorithms
are expected to withstand minor registration differences, but if
the source images are not registered, they should be subjected
to registration preprocessing steps which work independently
of the image fusion algorithm. The goal of an image fusion
algorithm is to accordingly combine similar and dissimilar
infonnation obtained from the source images in order to fonn a
new image which provides a more infonnative description of
the scene. The fused result can thereby be used to aid further
processing steps for a given task. J\s a result, image fusion
techniques are practical and essential for many applications,
including multi-spectral remote sensing [2], homeland security
[3], and biomedical imaging applications [4].
MULTI-SCALE IMAGE FUSION ALGORITHMS
A generalized multi-scale image fusion algorithm is
illustrated in Fig. 1. The input source images are transformed
using a given image decomposition technique. A fusion rule is
used to fuse the approximation coefficients at the highest
decomposition level. A different fusion rule is used to fuse the
detail coefficients at each decomposition level. The resulting
inverse transform yields the fmal fused result.
IMAGE FUSION ALGORITHMS USING THE PLIP MODEL
Adapting image fusion algorithms with the PLIP model
simply requires decomposition schemes and fusion rules to be
formulated in terms of the PLIP model. The combination of
the new PLIP images decomposition techniques with PLIP
fusion rules yields a new set of PLIP-based image fusion
algorithms. It should be noted that all PLIP decomposition
schemes are defmed for graytone functions. Therefore, images
are converted from images to graytone functions before PLIP
procedures are performed and converted from graytone
functions to images after PLIP procedures are performed.
CONCLUSIONS
In this paper, commonly used multi-scale image
decomposition schemes and image fusion algorithms were
combined with the PLIP model, yielding new image
decomposition schemes and image fusion algorithms.
Namely, the PLIP-DP, PLIP-DWT, and PLIP-SWT and
corresponding image fusion algorithms were proposed.
Furthermore, an analysis of the PLIP model provided a link
between the LIP and standard linear processing models, and
rationalized the use of the parameterized framework. Due to
the PLIP limit properties, LIP and standard linear processing
models were found to be instances of the PLIP model.