多层离散小波变换系数的稀疏向量构造的改进

融合离散小波变换和压缩感知的图像压缩方案很好避免了采用离散余弦变换和压缩感知时所带来的块效应，但当前基于单层离散小波变换的算法压缩比较低，基于多层离散小波变换的算法重构质量不佳。为了解决这些不足，根据离散小波变换系数的特点，对现有基于多层离散小波变换的算法提出了改进。图像经小波变换后，保留图像最高层低频系数，高频系数的构造方式给予适当改进。实验结果表明，与现有算法相比，重构图像的PSNR值得到2～4 dB提高。

An improved reconstruction to sparse vector for the coefficient of multilayer discrete wavelet transform

The image compression scheme which combines discrete wavelet transform and compressed sensing has overcome the block effect brought by the discrete cosine transform and compressed sensing. But the algorithm based on single discrete wavelet transform causes the low compression ratio, and the algorithm based on multilayer discrete wavelet transform causes a poor refactoring quality. In order to solve these problems, employing the characteristics of discrete wavelet transform coefficients, the better algorithm based on multilayer scheme is proposed. After the wavelet transform, reserving the low frequency coefficients at the highest level, the method to reconstruct the high frequency coefficient is appropriately improved. Compared with the existing algorithms, the experimental results of the proposed algorithm show that the PSNR of reconstructed image was improved about 2 ~ 4 dB.