Parallel Algorithms For Remote-sensing Image Fusion And Its Implementation

Remote sensing fusion technology is a way of making the most of the multi-sensor, multi-platform, multi-spectrum and multi-phase data. With the rapid development of sensor and platform technology, there are more and more approaches and sorts of the information we can get. As the result, the data sets gotten from the sensor are increasing tremendously. Meanwhile, fast processing technologies for many domains become more and more urgent, and the traditional single-processor can't achieve the requirement. However, the key to solve the problem is using parallel processing.Nowadays, the research point mostly is focusing on improving the fusion quality in China and the international world, and the research of parallel processing of the large number of data sets in remote-sensing fusion algorithms is in the beginning. So we do some research of parallel algorithms for remote-sensing image fusion which could make great progress in the efficiency and the speed of image fusion, which can be applied broadly.We analyze these requirements above in this paper, aiming at the research and implementation of three kinds of parallel algorithms for remote-sensing image fusion processing, and also implement the softwares for these parallel algorithms in web environments. The contributions and relevant work in the paper are as follows:Firstly, we discusses a new parallel algorithm of IHS image fusion――P-IHS, based on the discussion of the influence of four mainly RGB-IHS conversion models on the image fusion, and on drawing out the best judgment from the fusion image subjectively and objectively with the MS image and Pan image of the IKONOS. Then we analyze P-IHS image fusion method in theory.Secondly, we discusses a new parallel algorithm of PCA image fusion――P-PCA, based on the traditional PCA image fusion algorithm, and using the multinomial transform to improve the efficiency. We also test the parallel algorithm with the MS image and Pan image of the IKONOS on the pc clusters system with good results and high scalability.Thirdly, a parallel Discrete Wavelet Transform (P-DWT for short) fusion algorithm of remote sensing images is presented for the first time, facing the MIMD system architecture. To avoid the inherent communication overhead of the discrete wavelet fusion method, a special design called redundant partitioning is used, which is inspired by the characteristics of discrete wavelet transform.Finally, we implement RIPPS(Remote-sensing Image Parallel Processing System)software on the Browser/Server system, including many parallel algorithms of remote-sensing image processing. RIPPS is built on the thought of modularization and several level structures, and is easy for maintenance and scalability.