High-performance Remote-sensing Image Fusion Algorithm Research And It's Implementation

Remote-sensing image fusion is a processing that deals with multi-image data and other information from remote sensors. It focuses on using space (or time) redundancy (or complementary) multi-source data for computing according to certain rules to obtain more accurate and richer information than any single data and generate a new composite image with new spatial, spectral and time characteristics . With the developing of sensors, remote-sensing platforms and other related technologies, more and more ways to obtain remote-sensing information come forth, as well as the information style. At the same time, the amount of data obtained by means of remote-sensing is rapidly increasing, but many applications need more fast processing of remote-sensing image. Apparently, the traditional single-processor serial fusion algorithm couldn't meet the needs of such operations, and parallel is an efficient technical mean to solve this problem. Therefore, the research of parallel algorithm on remote-sensing image will greatly enhance the efficiency of image fusion, and it will have broad application prospects.We analyze these requirements in this paper, aiming at the research and implementation of there kinds of parallel algorithms for remote-sensing image fusion processing. The contributions and relevant work in the paper are as follows:Firstly, we analyze some algorithms choosing from three kinds of image fusion algorithms each, and on this basis we research and realize their parallelization. Then we test and verify the correctness of these parallel algorithms under real remote-sensing images and evaluate their performance.Secondly, we optimize the parallel image fusion algorithm concentrating to improve the way of communication and result write-back approach, reduce the communication time, improve parallel speedup and efficiency, and reduce the ratio of I/O time to the total running time.Thirdly, we research the fusion algorithm evaluation system. We first make an in-depth study on the impact of evaluation results under the existing indicators based on statistics or information entropy. After that we add performance standards into the existing evaluation criteria on fusion algorithm proposing an integrated quantitative evaluation model, and establishing a novel and easy quantitative evaluation system which is able to give consideration to both quality and efficiency.