A COMPREHENSIVE QUALITY EVALUATION METHOD OF INFORMATION FUSION
FROM HIGH-RESOLUTION AIRBORNE SAR AND SPOT5 IMAGES
Wenqing Dong, Qin Yan,
Chinese Academy of Surveying and Mapping, 100039 Beijing, P. R. China-(dongwq,yanqin)@casm.ac.cn
KEY WORDS: image fusion, SAR, SPOT5, entropy, Structural Similarity
ABSTRACT:
The image fusion of high-resolution Airborne SAR and SPOT5 can bring us high spatial resolution spectrum and texture information.
When analysing and evaluating the fusion image, we usually take the information entropy as a simple evaluation standard, which
maybe loss some other information we need. Actually, as many researchers have pointed out, it should be done comprehensively,
combining objective with subjective evaluation standards. That is to say, the optimum method is to carry on objective quantitative
evaluation on the basis of subjective qualitative visual evaluation. According to the image characteristics of airborne SAR and
SPOT5, this paper discussed several indexes such as entropy, mean value, standard variance and so on. And, combining with
subjective analysis result, a comprehensive quality evaluation of the fusion images with structural similarity can be obtained. At the
last, a test has testified it and corresponding analysis result has been presented.
1. INTRODUCTION
The fusion technique of remote sensing images aims to
improve the capability of application and detection the ground
objects. And usually, the fusion image data is a group of new
space information and combined images. So that, for the
different sensor data, it can combine the advantages of different
remote sensing data resources, compensating the deficiency of
some data resource, to decrease the uncertainty, improve the
accuracy of remote sensing image classification and the
capability of dynamic monitoring (Gemma, 2004).
During the recent years, the quick development of Synthetic
Aperture Radar (SAR) got more and more attention in the
world. SAR has the advantage of full-time, full-weather,
penetrability for cloud and rain, even some vegetation with the
change of radar wavelength, high-resolution (e.g. airborne SAR
with 1-meter resolution), and more texture information. But in
practical applications, there exists some type of the ground
objects which are difficult to be recognized by human eye’s
vision, as a result of SAR side-imaging mode, the influence of
noise and shadow, vision levels and so on. For example,
because of the shelter and shadow, it will be difficult for the
trees and buildings with height to produce mapping quickly
(Guo, 2000; Ulaby, 1987).
At the same time, SPOT5 image data has the characteristics of
high resolution, capability of side observation to produce stereo
images, and obtaining image of the same area in short time,
which are gained more and more attention from the consumers
in remote sensing field. In the research fields such as land use
and managing, forest cover monitoring, soil eroding, and city
planning, it has shown important effect. So, the fusion image
can take full advantage of SAR and SPOT5 data, to obtain the
multi-level information of ground objects, improve the objects
characteristic and the classification accuracy of the ground
cover (Pohl, 1998; Tupin, 2003; Solberg, 1997). 2
2. THE DECISION OF OBJECTIVE AND
SUBJECTIVE EVALUATION INDEXES FOR THE
FUSION IMAGES OF SAR AND SPOT5
2.1 Use of Fusion Image and Evaluation Objective
In this paper, the fusion image of high-resolution airborne SAR
and multi-spectral SPOT5 data is mainly applied in terrain
mapping, so that, there will be three objectives for the image
fusion as following that are:
• To improve image resolution: Generally speaking, the
processing of image fusion between the lower-resolution multi-
spectral remote sensing data and the higher-resolution airborne
SAR data can obviously improve the image resolution,
simultaneously preserving enough texture information.
• To enhance the information quantity: For an effective
image characteristic extracting, it is necessary to enhance the
information quantity of ground types from SAR data, to obtain
the facility of vision recognizing.
• To improve the definition: During the image processing,
we need improve the image quality enhances the image detail
information, and texture characteristic, under the condition of
preserving the original image information as soon as possible.
For SAR image, it is difficult to preserve the edge information
and energy utilizing the common ways of image enhancing
unless the image fusion.
2.2 Choice Method of Evaluation Index
Usually, for the image quality evaluation of fusion images, the
evaluation index items applied in fusion image evaluation are
included the subjective and objective evaluation indexes,
decided by the objectives and use of the fusion image from
high-resolution airborne SAR and SPOT5 data. The subjective
evaluation indexes are often the base criterion for the image
interpreting researchers, with simple and direct characteristic.
Meanwhile, the common objective index items involve such as
information entropy, mean gradient, correlation coefficient,
and mean value, standard variance and so on. However, in the
processing of evaluating the fusion images, there exist many
subjective factors influencing the eventual evaluation result,
and the common objective evaluation methods may not make
the best of the characteristic of human visual system (Wang,
2003).
In this paper, the choice of evaluation index will mainly
depend on the relationship between the original and fusion