中文摘要：

研究高光谱图像的稀疏分解匹配优化问题,为便于对图像进行压缩处理,但正交匹配追踪算法的计算复杂度非常高,难以用于实时处理。针对高光谱图像,提出采用粒子群优化的图像稀疏分解算法,对正交匹配追踪算法的匹配过程进行优化,依靠粒子群算法的局部寻优能力,快速找到最优原子,完成图像稀疏分解。实验结果表明,在构造的Gabor冗余字典基础上,改进算法得到的重构图像峰值信噪比能达到44dB以上。同时,与正交匹配追踪算法相比,上述算法计算复杂度低,计算效率提高14倍,且算法不需要事先产生冗余字典,减少对存储空间的占用,满足实时性要求。

英文摘要：

Sparse decomposition of hyperspectral imagecan facilitates image compression. However, the computational complexity of orthogonal matching pursuit （OMP） algorithm is too high to be applied to real-time processing. For hyperspectral images, the paper proposed an image sparse decomposition algorithm based on particle swarm optimization （PSO）, to optimize the matching process in OMP. The best atoms were found quickly with the local optimization ability of PSO to realize the sparse decomposition. Experimental results show that based on the constructed Gabor redundant dictionary, the reconstructed peak signal-to-noise ratio can hit 44 dB or more. Meanwhile, compared with OMP, the proposed algorithm can reduce the complexity of sparse decomposition and increase the computational efficiency by 14 times. Since the proposed algorithm does not need to construct the redundant dictionary, the storage space occupied by redundant dictionary has also been reduced to meet real-time requirements.