The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part BI. Beijing 2008 
3. RESULTS ON RADIOMETRIC PERFORMANCE 4. RESULTS ON GEOMETRIC PERFORMANCE 
Radiometric image quality is not raised as a concern and not 
commented upon by many authors. Armenakis & Beaulieu 2006 
state that image quality is good except in the forward end of 
Band F, where it is less sharp. Gachet & Fave 2006 and 
Baltsavias et al 2007 comment on low dynamic range as a 
major handicap for matching process after converting the 10-bit 
data set to 8-bits due to algorithmic/software constraints faced 
by them. However there are comments to the opposite by 
Jacobsen 2007 as to the contrasting details that can be observed 
even within snow-covered fields. Armenakis & Beaulieu 2008 
report that the dynamic range is good, making terrain features 
clearly visible and terrain morphologies differentiable. Lehner 
et al 2006 state that MTF of the aft-looking sensor is much 
better than the MTF of the fore-looking sensor by pure visual 
inspection. Kay & Zielinski 2006 comment on the image quality 
as very suitable for DSM generation given the sub-optimal 
image acquisition date (Jan.) for the Mausanne test site. 
Baltsavias et al 2007 also comment on the difference in image 
sharpness and scale differences due to shadows between Fore 
and Aft images. As an exceptional case, they also comment on 
having observed artifacts, interlacing errors and pattern noise in 
Aft image of Rome scene after converting the 10-bit data to 8- 
bits and some preprocessing. Baltsavias etal 2007 also observe 
horizontal edge jitter in Fore image of Rome scene. 
Since these observations have been made in 2006 and 2007, R. 
Nandakumar etal 2008 report on the improvements carried out 
to the Cartosat-1 orthokit products with regard to improving the 
MTF after applying a scene-based stagger correction and 
obtaining improved results in terms of both visual quality as 
well as improved DSM derivation. 
The results of stereo image matching also speak of image 
quality, in an indirect way. In Mausanne Jan scene 81.5% of the 
points matched had correlation coefficient better than 0.6 as per 
Kay & Zielinski 2006. Jacobsen 2006 reports 84% for the same 
scene and 93% for the Feb. acquired scene. For the snow 
covered Warsaw test site, J. Zych et al 2006 report 84% 
correlation with correlation coefficient greater than 0.8. 
Jacobsen 2006 reports 94% correlation with correlation 
coefficient greater than 0.6. As per Lehner et al 2008, the 
number of conjugate points identified after initial pixel level 
correlation, followed by least squares matching for sub-pixel 
identification and supplemented with region growing with built- 
in blunder detection checks result in 7.08 million points for the 
Catalonian stereo pair of 12 k by 12 k pixels and 4.82 million 
for the Jan scene over Mausanne and 6.14 million for the Feb 
scene over Mausanne. Lehner et al 2008 report constructing a 
DSM of 5 m grid spacing for the Catalonian stereo pair and a 
DSM of 10 m grid spacing for the Mausanne stereo pair. 
According to Armenakis & Beaulieu 2006 & 2008, the 10-bit 
dynamic range enables the detection and identification of 
features and terrain patterns such as roads and 
geomorphological patterns, as they are visible and differentiable 
according to the 2.5 m spatial resolution. B. Sadasiva Rao et al 
2006 report the ability to extract several types of vector features 
from Cartosat-1 Aft orthoimages, although not over C-SAP test 
sites, which are given in Table-2. 
4.1 Choice of Stereo Angles for Cartosat-1: Gruen 2008 
reports that despite the quasi-simultaneous image 
acquisition, 
Features 
Feature Class 
Cultural 
features(polygon) 
Buildings, Group of Buildings, Parks, 
Play Grounds, Swimming Pools, 
Stadia. 
Transportation (line) 
Metalled roads, Unmetalled roads, 
Bridges, Culverts, Flyovers, Lane, 
Footpaths, Railway Lines, and 
Traffic Island (polygon) 
Vegetation 
Single (point), Grove (polygon), and 
Plantation (polygon). 
Hydrography 
(polygon features) 
Water filled river, Dry river, Water 
filled and Dry Streams, Drains. 
Hydrography (point 
features) 
Embankments, Overhead tanks, 
Ground level reservoirs. 
General (polygon) 
Marshy lands, Rocky areas, Scrub 
lands, and Quarry sites. 
Table-2: Culturable features mappable from Cartosat-I 
the two images show often radiometric differences that lead to 
measurement errors. This is partly due to the unfavorable choice 
of viewing angles for the Aft and Fore channels, which also 
leads to scale differences between the images, causing errors in 
matching. However, Lehner et al 2008 report that quote: “The 
numbers of tie points found and their sub-pixel accuracy is 
highly dependent on the stereo angle. A large stereo angle (large 
base to height ratio b/h) leads to poorer numbers of tie points 
and to lower accuracy in LSM via increasing dissimilarity of 
(correctly) extracted image chips. For currently available high 
resolution stereo imagery the stereo angle is too large, at least 
for built-up areas. The importance of a large base-to-height ratio 
is exaggerated at the cost of the matching accuracy and density 
(see Krauss et al, 2006). The accuracy in forward intersection is 
inversely proportional to the base-to-height ratio but also direct 
proportional to the matching accuracy. The latter and the 
matching density are improved by reducing the stereo angle.” 
unquote. Jacobsen 2007 in conclusion says: quote: “The stereo 
models of Cartosat-1 have optimal conditions for the generation 
of digital height models by automatic image matching. The 
short time interval between both images avoids a change of the 
object and shadows between imaging. The height to base 
relation of 1.6 is a good compromise for open and not too dense 
build up areas. A larger angle of convergence often causes 
problems in matching especially in mountainous and city areas, 
so the percentage of accepted matched points may be smaller 
than the reached 84% up to 94%. On the other side a smaller 
angle of convergence has a negative influence to the accuracy 
but advantages for city areas. With a standard deviation of the 
x-parallax between 0.49 and 0.80 GSD similar x-parallax 
accuracies like with the comparable SPOT HRS have been 
reached (Jacobsen 2004). Of course with the different GSD and 
different height to base relation the absolute vertical accuracy 
based on SPOT HRS cannot be as good like for Cartosat-1. Of 
course the matching results depend upon the used area. In 
general open areas with sufficient contrast are optimal, but also 
under the not so optimal conditions of forest the achieved 
results are satisfying.” Unquote. 
1281