基于ι~∞范数的稀疏ICA和FCM的自然图像特征提取

Image feature extraction based on sparse ICA with ι~∞ norm and FCM

为了保证图像特征系数的稀疏性和加快寻找最优基的收敛速度,提出了一种基于ι∞范数的稀疏独立分量分析(SICA)的算法。该SICA算法采用ι∞范数作为ICA的稀疏性度量标准,用模糊C均值聚类算法初始化独立分量的特征基,有效地实现了自然图像的特征提取;而且,该SICA方法不需要优化高阶的非线性函数和密度估计,因而计算简单、且收敛速度较快;同时,利用提取的图像特征成功地实现了图像恢复,通过图像恢复对比实验表明了该方法在特征提取方面的合理性和实用性。

To ensure the sparsity of image feature coefficients and improve the convergence speed of finding the optimized features,a method of sparse independent component analysis（SICA） based on ι~∞ norm is proposed.Features of natural images is extracted efficiently by this SICA algorithm,which utilizes the ι~∞ norm as the sparse measurement criterion of ICA and the fuzzy C-mean clustering（FCM） method to initialize feature basis vectors of independent components.Moreover,this SICA method does not need optimizing the high-order non-linear functions and density estimation,therefore,it is very simple in computing and its convergent speed is also very quick.At the same time,the image restoration work is successfully implemented by using these features extracted.Further,compared with other methods of image restoration,the experimental results testify that this feature extraction is also reasonable and practical in application.