基于免疫克隆的图像稀疏分解算法研究

Research on sparse decomposition algorithm of hyperspectral image based on immune cloning

图像的稀疏分解能够得到信号的简洁表达形式,提高后续处理的效率。采用冗余字典对图像信号进行稀疏表示时,更容易捕捉到信号的特征。文中构建了谐波小波包冗余字典,采用免疫克隆的生物进化思想,利用免疫克隆、克隆变异、克隆选择的操作算子完成抗体的更新进化,以寻找到最优原子,并利用最优原子对原始图像进行稀疏表示。采用Matlab软件进行仿真,对高光谱图像和一般二维图像进行稀疏分解。实验结果表明,谐波小波包字典得到的重构图像的峰值信噪比高出Gabor冗余字典2 dB以上,且重构图像更加光滑,充分说明谐波小波字典具有更强的稀疏表示能力。

The sparse decomposition of the images can obtain a concise expression form of the signal and improve the efficiency of subsequent processing.When the redundant dictionary is used to sparsely represent the image signal,it is easier to capture the characteristics of the signal.The paper constructs a harmonic wavelet packet redundancy dictionary,adopts the biological evolution idea of immune cloning,and uses the operators of immune cloning,clonal mutation,and clonal selection to complete the update and evolution of antibodies to find the optimal atom,and then uses the optimal atoms to sparse represent the original image.Using Matlab software for simulation,the hyperspectral image and general two⁃dimensional image are sparsely decomposed.The experimental results show that the peak signal⁃to⁃noise ratio of the reconstructed image obtained by the harmonic wavelet packet dictionary is 2 dB higher than that of the Gabor redundancy dictionary,and the reconstructed image is smoother,which fully demonstrates that the harmonic wavelet dictionary has stronger sparse representation ability.