Research On A Rapid Fusion Method For Remote Sensing Images Based On An Improved Atrous Wavelet Decomposition

From the beginning of the 1970 s, China showing a vigorous development trend in the aerospace, space and remote sensing information technology. Satellite remote sensing has played an important role in the military and economic construction which is a branch of remote sensing. Since the "Twelfth Five-Year", China's civil aerospace industry put forward a new development goals and direction. The development of satellite remote sensing has reached a new height.With the development of remote sensing,more and more satellites have been come into use. The image which get from a single sensor has showed some limitations in spatial resolution and spectral information, so it is necessary to fusion the images which get from different sensors. On the other hand, as the amount of remote sensing data increases, the way to increase the processing speed which by enhance the hardware performance had run into the bottleneck. The image's timeliness which get from the traditional computing model couldn?t satisfy current demand.To solve those problems, the following research was made:1. In this article, the difference and characteristic of commonly used fusion algorithms is introduced. Then it analyzed the advantages of the fusion algorithm which based on atrous wavelet decomposition and made a comparative study of atrous wavelet decomposition algorithm and other fusion algorithm on the parallel processing. For the loss of fusion result's details in wavelet decomposition process, this article proposed a new image fusion algorithm which based on high-frequency emphasis and atrous wavelet decomposition. This new fusion algorithm use high-frequency to strengthen the wavelet planes which get by atrous wavelet transform, and it verified by experiments proved that this algorithm can reserve more detail information of images which have a complex topography.2. It depth analyzed the CUDA parallel architecture's structural characteristics, features and design principles. On the base of this analysis, this paper proposed a parallel optimization strategy which based on GPU and created a new parallel fusion method of remote sensing. Then this article specifically describes the way how to use CUDA parallel architecture to improve the improved algorithm's computational efficiency and then demonstrated the core code. Finally, the efficiency of serial computation and parallel computation were compared by experimental, and it proved that the efficiency of the improved algorithm is higher than the original algorithm.