Adjusting astronomical images using a censored Hausdorff distance

We present in this paper a fast method to find a translation between a pair of astronomical images. Being able to match stars very precisely is necessary in numerous problems in astronomy. The method we describe here can be used to get a first estimate with a precision of a few pixels. It can be used to initiate a second method to achieve the desired sub-pixel accuracy. Implementation on a massively parallel computer has been made, showing that real-time is feasible, and the development of a specific hardware looks promising.