FUSION OF THEMIS AND TES FOR ACCURATE MARS SURFACE CHARACTERIZATION
C. Kwan, B. Ayhan, and B. Budavari
Signal Processing, Inc.
ABSTRACT
This paper presents a novel approach to fusing Thermal
Emission Imaging System (THEMIS) and Thermal Emission
Spectrometer (TES) satellite images, aiming to improve
Mars surface characterization performance from orbit. Our
approach includes proven registration and advanced
pansharpening algorithms developed by us and others.
Preliminary experiments show that the fusion approach is
highly promising despite the extremely high resolution
difference of THEMIS and TES (30 to 1). We also observed
some potential issues that require further research.
Index Terms— image fusion, pansharpening, THEMIS,
TES, registration
1. INTRODUCTION
Recently, NASA has expressed strong interest in improving
surface characterization of Mars using orbital imagers.
THEMIS [1] and TES [2] are orbital multispectral imagers
of Mars. THEMIS has 10 spectral bands in the 6-13
micrometers region and a spatial resolution of 100 m. TES
has 143 spectral bands in the 5-50 micrometers range, but
with low spatial resolution of 3x3 km. Although both have
been used to map out the surface composition of Mars, there
are some limitations. First, THEMIS has low spectral
resolution that may not provide accurate surface
characterization. Second, TES has low spatial resolution that
cannot provide fine spatial details of surface characteristics.
Fig. 1 illustrates the spatial difference between THEMIS and
TES images. Roughly speaking, each TES pixel contains
about 900 THEMIS pixels. It is therefore very challenging
to fuse the two data sets.
There are 4 research questions that we would like to
address:
Q1: Image registration: In addition to aligning THEMIS and
TES bands, it turns out that THEMIS bands also have
large shifts. How can one compensate for the shifts
observed in the THEMIS bands?
Q2: Pansharpening performance: Normally, pansharpening
is only done for fusing images with 3 to 4 times
resolution differences. Will pansharpening algorithms
still perform well despite the huge resolution difference
(30:1) between THEMIS and TES?
Q3: Data compatibility between THEMIS and TES:
Although both THEMIS and TES are thermal imagers,
do they have the same signal response?
Q4: Surface characterization from orbit: What is the
achievable and quantifiable performance using orbital
imagers as compared to in situ and ground based
instruments?
Fig. 1: Spatial resolution difference between THEMIS and
TES. Background shows one band of THEMIS. Each dark
square shows a single TES pixel.
In this paper, we present the first application of a novel
approach to fusing THEMIS and TES images. The goal is to
eventually improve Mars surface characterization from orbit.
It will be a long way to get there. We believe the first step is
to fuse THEMIS and TES to form a high spatial resolution
and high spectral resolution image cube. After that, we can
then apply advanced hyperspectral image processing
algorithms to analyze the pixels in the data cube. We are
currently focusing on answering Q1 and Q2 and have
observed some issues for Q3. Q4 will be addressed in the
future. The fusion consists of two steps. First, the THEMIS
and TES images need to be aligned. Even within the
THEMIS bands, we observed that the bands have serious
misalignments. To align the THEMIS bands, we extracted
Speeded Up Robust Features (SURF) [3] and then applied
RANSAC (Random Sample Consensus) technique [4]. To
align THEMIS and TES, we utilized the coordinate
information stored in the metadata of the images. Second,
we evaluated 10 pansharpening algorithms to the registered
image cube using a synthetic data. Due to the huge
resolution difference between THEMIS and TES, some
algorithms worked and some did not. Both objective (4
performance metrics) and subjective evaluations (false color
3381 978-1-5090-4951-6/17/$31.00 ©2017 IEEE IGARSS 2017