Least-squares performance of analog product codes

In this article the least squares method is utilized for the decoding of analog product codes. In particular it is shown, (i) how the LS solution is constructed for arbitrary image block sizes, (ii) that the theoretical achievable SNR improvement after LS decoding is inverse proportional to the code rate, and finally (iii) the LS results are compared with a known iterative method. The convergence of the iterative method is analyzed, allowing for optimal behavior and a consistent proof of convergence is provided. Theoretical results are compared to simulations on the example of a picture transmission.