The GPGPU Application In Remote Sensing Image Fusion

Along with the computing capability of Graphic Processing Unit (GPU) getting more powerful, GPU is widely applied to the general purpose computing not just restrict to graph manipulation. Remote sensing image data can be parallel processed in many of the image fusion algorithms, this paper utilizes the GPU character that parallel operation on the data, focuses on the GPU implementation of the remote sensing images fusion in spatial domain and transformed domain. The main contents and contributions in this thesis can be summarized as follows:1,The method of generating floating-point textures of different formats on GPU is analysed. In order to utilize the parallel processing characteristic in data-level of remote sensing image fusion algorithm on GPU, Three ways of data packing are proposed and the effect of each means is analysed on the computing speed on GPU.2,The fusion algorithms in spatial domain are mapped to the SIMD computing way on GPU, the Cliche product fusion, Brovey ratio fusion, high-pass filtering fusion and weighted fusion are realized using GLSL, and the speed of the fusion is accelerated using RTT (Render to Texture) technology. The texture-lookup time is reduced in fragment shader to accelerate the computing speed by creating assistant textures in high-pass filtering fusion and HVS weighted fusion.3,A IHS transform algorithm based on GPU is proposed in this paper and the IHS forward transform and inverse transform are mapped to two fragment shading processes, parallel computing and outputting of the 3 component in both transform processes is realized using MRT (Multiple Render Targets) technology. The data exchange between forward and inverse transform is realized by changing source buffer and destination buffer to save the memory and improve the computing efficiency.4,A DWT (Discrete Wavelet Transform) fusion algorithm is designed in the paper, 2D DWT is divided into two steps of 1D DWT. A indirect address texture is created for every transform step and the transform of each level is based on the result stored on the texture of the last level. A FBO is set for every image to be fused to restore intermediate data an to do data exchange. The computing process is completely implemented on GPU to avoid the data exchange between GPU and CPU. The experiment shows that the speed of the DWT fusion on GPU can be more than 10 times than that on the CPU.