Image registration is an important step of image fusion

change detection and target recognition in remote-sensing applications. As a similarity measure

mutual information has become a hot topic in image processing

for its advantages of no pre-processing

robustness and a high degree of automation. But with the increasing of image size in remote-sensing

high computing complexity of image registration algorithm makes that traditional single-processor computing mode could not meet the requirement of real-time processing in some applications. Based on the experimental analysis of corresponding serial algorithm

a parallel region-based registration algorithm based on mutual information for remote-sensing images is proposed. Strategy of data partition and parallel computing method of mutual information are given. To avoid data communication

some techniques like redundant data partition on boundary and reducing on binary-tree are used. Then

a quantitative analysis of computing difficulty is given. The experimental results show that the proposed parallel algorithm has good scalability and applicability.