The convergence rate of iterative algorithm for image reconstruction can be accelerated by ordered subsets method

but the image quality degrades due to lack of statistical information within subsets.An approach of subset partition is proposed based on hypothesis test.Two test statistics are established and the iterative formula is given.The method can automatically adjust the number of the subset for each iterative according to the significant level demanded by user and form the subset with the same statistical information content.The experimental results demonstrate that this algorithm can converge faster and provide high-quality reconstructed images after a few iterations.