A proposal of parallel strategy for global wavelet-based registration of remote-sensing images

With the increasing importance of multiple multiplatform remote sensing missions, digital image registration has been applied into many fields, and specially plays a very important role in remotely sensed data processing. Firstly a brief introduction of existing parallel methods of wavelet-based global registration is given. And then the communication optimization for GP method is described. The optimized algorithm is named Group-Optimized-Parallel (GOP for short). To find out the reason of occasionally lower efficiency of GOP than other methods, a more careful analysis is presented in theory and proved in experiments. Moreover, we give a quantitative criterion, called Remainder Items, to choose the best solution in different input conditions.