A jump and look all round long range image restoration and its parallelism

An image information restoration algorithm based on long-range correlation (LR-IIRA) can be efficiently applied in image interpretation, restoration and error concealment. However, this algorithm requires excessive computing time, which restricts its application to large-scale problems and real-time problems. We propose an improved LR-IIRA, named jumping and looking all round (JL-LR-IIRA). By using this algorithm, the computing time can be reduced to 1/6 of LR-IIRA's, i.e., the computing time is 14.17 (seconds) by LR-IIRA and 2.41 (s) by JL-LR-IIRA. To reduce the processing time further, a parallel JL-LR-IIRA algorithm (PJL-LR-IIRA) to parallelize the restoration processing is proposed. In our PJL-LR-IIRA, after we analyze data distribution and communication deeply, a parallel read method and a key position technique are presented. By using our PJL-LR-IIRA on four computing nodes on the Dawning 2000 cluster system, the processing time is 0.69 (second). PJL-LR-IIRA provides a new efficient way for real-time image restoration applications. Furthermore, from theoretical analysis and experiment results, it shows that the scalability of our PJL-LR-IIRA is good. So PJL-LR-IIRA is also a new efficient way for large-scale image restoration applications.