1055 
AUTOMATED NOISE REDUCTION IN HRSC MARS DTMS 
Shih-Yuan Lin*, Jung-Rack Kim, Jan-Peter Muller 
Mullard Space Science Laboratory, University College London, Holmbury St. Mary, Dorking, RH5 6NT, UK - 
(syl, jk2, jpm)@mssl.ucl.ac.uk 
KEY WORDS: Extra-terrestrial, Mapping, DTM, Surface, Matching, Geology, Noise, Detection 
ABSTRACT: 
Since the European Space Agency Mars Express spacecraft entered an orbit around Mars, the High Resolution Stereo Camera 
(HRSC) placed onboard has been imaging the Martian surface. Due to the characteristics of high resolution and stereo imaging 
capability of the HRSC, the construction of high resolution digital terrain models (DTMs) representing Mars surface topography has 
been achieved. However, during the process of DTM creation, errors will occur which may lead to incorrect interpretation. 
Therefore, the certification of DTM quality is required before any geological investigation is performed. This paper describes a new 
method to detect and remove noise existing in Martian terrain models produced from HRSC stereo imagery. The proposed method is 
here demonstrated using HRSC hi011 orbital images and positive results were achieved, in which the number and position of the 
noisy heights were extracted. Furthermore the method proved of great potential in reducing vertical discontinuities between adjacent 
DTMs when DTM mosaicing was performed. This solution for automated noise reduction has demonstrated the capability for 
refining Mars DTMs produced from HRSC stereo imagery, which will be helpful in improving hydrological analysis and geological 
investigations of the Martian surface. 
1. INTRODUCTION 
A fundamental scientific question in the exploration of the 
planets within our Solar System is whether conditions 
favourable to the development and presence of life exist or have 
existed in the past. Central to this issue is the search for 
evidence of past (and present) water, including the 
determination of how much water was present at or near the 
surface. Of all the planets in the Solar System, Mars is perhaps 
the best (and nearest) candidate for addressing these questions, 
and hence is the focus of significant planetary exploration at 
present. Determining the inventory of water reservoirs and 
morphological parameters on outflow channels on Mars 
provides critical information for understanding the geological, 
climatic and potentially exobiological evolution of the planet. 
However, previous studies of the role of water on Mars have 
been limited by the spatial resolution of topographic data 
available to study Martian landforms. Whilst extensive but low 
resolution Viking and MOC-WA (MOC-Wide Angle) imagery, 
and high resolution, but spatially restricted, imagery taken from 
MOC-NA (MOC-Narrow Angle) and Thermal Emission 
Imaging System (THEMIS) have been available, the lack of 
high resolution topographic data with which to investigate 
meso-scale (10m-100 km scale) landforms in detail has 
restricted the analysis and interpretation of Martian surface 
features. A detailed investigation of landscape formation and 
evolution requires high resolution (<<200m) and accurate 
topographic data (Zrm<30m) from which fine-scale clues to the 
origin of landforms can be resolved and quantitative 
measurements of landscape properties determined. 
Since the European Space Agency (ESA) Mars Express 
spacecraft entered an elliptical orbit around Mars on the 25th of 
December 2003, the High Resolution Stereo Camera (HRSC) 
placed onboard has been imaging the Martian surface. Due to 
the characteristics of high resolution and stereo imaging 
capability of the HRSC, the construction of Digital Terrain 
Model (DTM) of Mars surface topography with grid spacing of 
30-75 metres is feasible (Albertz et al., 2005; Heipke et al., 
2007). By visualising the Martian terrain in a high resolution 
three-dimensional model, the evolution of the Martian surface 
and the geological processes involved can be effectively 
analysed. 
In order to select an area for future exploration, DTM data 
quality is a very important aspect of the geological investigation 
of the Martian surface. Errors/blunders and artefacts occurring 
in any DTM will lead to incorrect interpretation and 
understanding of the underlying physical processes. This paper 
therefore aims to detect and remove such noise existing in any 
Martian terrain model produced from HRSC stereo imagery. 
Moreover, we deal with errors appearing in the final DTM 
product, rather than the ones generated in the intermediate steps 
in the process of DTM generation. 
The strategy of noise reduction follows a two-stage process. 
Surface matching is the technique first applied to determine 
noisy DTM points and remove them in the DTMs. 
Subsequently, the points removed are further examined using 
the corresponding pixels in the ortho-image. Their colour digital 
number (DN) values are inspected to determine if there is any 
point being removed improperly. After extensive testing, the 
overall solution for automated noise reduction, proposed in the 
paper, is shown to be an efficient method for refining Mars 
DTMs produced from HRSC stereo imagery. The method is of 
great potential value for improving hydrological analysis and 
geological investigations of the Martian surface. The details of 
the method, as well as its implementation, results, quality 
* Corresponding author.