一种多传感器图像并行融合新方法

小波包变换能够为图像融合提供非常精细的分析。但是,在图像较大时融合的计算量大,耗时长,难以进行快速、实时融合。通过对小波包融合方法在单处理机上的执行效率的分析,针对算法固有的时间复杂性和并行性,提出了一种分布存储环境下的小波包并行融合算法。该算法针对小波变换计算的数据局部性特点,设计并实现了基于Pentium PC和1000Mbps交换式以太网的机群系统的MPI(Message Passing Interface)并行环境的并行小波包图像融合方法。针对不同大小图像、以及不同的集群规模,分析了并行计算的性能。在机群系统上对算法进行实现,结果表明该算法具有良好的并行性能。

A New Parallel Multi-sensor Image Fusion Method

To enhance the information from different sensors and improve the analytic and extracting ability of images,an algorithm of image fusion based on the wavelet packet transform is proposed after studying the wavelet packet analysis method.Although the analyzing method based on the wavelet packet is more accurate than that based on the multi-resolution wavelet,with the former method,computing complexity is exponentially increased with the growth of image size.After analyzing the efficiency of image fusion algorithm based on the wavelet packet on the single computer,the inherent complexity and the parallelism of the above fusion algorithm,a parallel image fusion based on wavelet packet in the MPI(message passing interface) framework is proposed.In addition,the efficiency of parallel computing is studied under different conditions as different image size and different cluster size.The experimental result shows that the efficiency of parallel computing is obviously improved and this paves the way for the future design of highly efficient and real-time fusion algorithms.