AN UNBIASED AND COMPUTATIONALLY EFFICIENT LS ESTIMATION METHOD FOR IDENTIFYING PARAMETERS OF 2-D NONCAUSAL SAR MODELS

This paper presents an unbiased and computationally efficient modified least squares (LS) estimation method for identifying parameters of two-dimensional noncausal simultaneous autoregressive models. Some intuitive analysis and a mathematical proof of the unbiasedness of the new LS estimation are given, and a recursive in-order fast algorithm to implement the new LS estimation is introduced. Finally, computer simulation results are given to sustain the theoretical analysis in the paper. Both the theoretical analysis and the computer simulation show that the new LS estimation possesses much higher estimation accuracy and lower computational complexity than those of the conventional LS estimation method. Comparing with the approximate maximum likelihood method by Kashyap and Chellappa, our scheme is much faster with the same estimation accuracy and yet parallelizable.