摘要

The most popular iteration method used in Direct Demodulation Method (DD) is Richardson- (RL) Iteration. The formula of RL iteration can be rewritten in matrix form. There are two matrix multiplications which contain the main computation cost. They can be transformed into convolution if the system is shift-invariant. As is well known, convolution can be computed by Fast Fourier Transform much faster than computed directly. This paper introduces the details of the above procedure, which is called accelerated direct demodulation method (ADD), and applies the procedure to image restorations of Hard X-ray Modulation Telescope data. This paper also compares the computation cost between the original DD and ADD.