A unique vision-based three-axis spacecraft attitude determination
method using Moon or Earth observation images is introduced,
which is able to operate in a stand-alone configuration.
We describes the image data processing algorithms to extract the
two directions needed from the observed images to estimate the
spacecraft orientation. Since the image rotation and translation processes
affect the attitude estimation result, the accuracy is quantified
by analyzing the sensitivity of the two directions due to subpixel
image translations and rotations. Furthermore, with the vision of a
new Mars mission, an alternative image data processing algorithm
for oblate bodies is presented. As for the attitude accuracy level that
can be achieved, it highly depends on the image processing
efficiency/complexity. Its numerical value, however, is not quantified
because it would require unavailable real data.