基于广义拉普拉斯分布的图像压缩感知重构

来源期刊:中南大学学报(自然科学版)2013年第8期

论文作者:何宜宝 毕笃彦

文章页码:3196 - 3203

关键词:图像压缩感知; 先验分布; 广义拉普拉斯分布; 非凸优化

Key words:image compressed sensing; prior distribution; generalized Laplacian distribution; nonconvex optimization

摘    要:针对图像压缩感知重构问题,构建图像小波系数的广义拉普拉斯统计模型。首先通过对典型图像小波系数的直方图统计,以广义拉普拉斯分布拟合图像小波系数的先验概率密度,用KL散度法求得广义拉普拉斯分布的参数。然后基于贝叶斯准则,通过取对数,将稀疏系数的最大后验概率估计问题转化为p范数优化问题,其中p的取值由待重构的图像所决定,即为该图像小波系数对应的广义拉普拉斯分布的形状参数。最后由非凸优化法求解得到图像的小波系数,并实现图像的重构。实验结果表明:对于简单稀疏信号,该方法重构成功率明显高于经典的BP和OMP法;对于测试图像的小波系数信号,所提方法能够自适应地精确重构原始图像。

Abstract: A probabilistic sparse model based on generalized Laplacian distribution is constructed to solve the popular problem of image compressed sensing. Through histogram statistics of wavelet coefficients of typical images, the generalized Laplacian distribution is chosen to fit prior distribution of the coefficients. Parameters of the distribution are estimated with a method of KL divergence. Based on Bayes’ rule, estimation of the maximum a posteriori of coefficients is transformed to a minimization problem of p norm, in which, the value of p is determined by shape parameter of the generalized Laplacian distribution corresponding to current image. Using nonconvex minimization, the coefficients are reconstructed. Experimental results show that, for simple sparse signal, exact reconstruct frequency of the proposed method is higher than BP and OMP; for the wavelet coefficients of test images, the proposed algorithm can reconstruct image exactly and adaptively.

相关论文

  • 暂无!

相关知识点

  • 暂无!

有色金属在线官网  |   会议  |   在线投稿  |   购买纸书  |   科技图书馆

中南大学出版社 技术支持 版权声明   电话:0731-88830515 88830516   传真:0731-88710482   Email:administrator@cnnmol.com

互联网出版许可证:(署)网出证(京)字第342号   京ICP备17050991号-6      京公网安备11010802042557号