A new method for fast shift determination

Abstract

We describe a new, fast method for shift determination. The method is based on a binary code vector and a shift matrix in which all possible shifts of an input signal are stored. For determination of shifts of an 1D real signal with N pixels this method needs at first N/2 operations of comparison to get the binary code vector with N/2 components, and it will then be converted into N/m m-bit integer numbers which indicate where the correct shift is stored in the shift matrix. For N=256, one needs only about 6 ms by our host computer using a 68000 processor for the shift determination. In addition, this method is stable against the random noise and a loss of part of the signal. Furthermore, this method is invariant under the change of contrast β · a + b where a is an input signal, β is a positive amplification, and b a possible bias of the signal.