论文部分内容阅读
提出了一种小波系数的高斯混合分布模型。该模型用两个分量的高斯混合分布来拟合小波域每个高频频带系数的分布:其中,先验概率小方差大的分布代表了少量幅值较大的小波系数;先验概率大方差小的分布代表了大量幅值较小的小波系数。该文采用基于特征的快速算法完成对高斯混合分布参数的求解。实验结果证明了模型的有效性。
A Gaussian mixture distribution model of wavelet coefficients is proposed. In this model, the Gaussian mixture distribution of two components is used to fit the distribution of each high-frequency band coefficient in the wavelet domain. Among them, the distribution with small variance of small priori shows a small number of wavelet coefficients with large amplitude; A small distribution represents a large number of smaller wavelet coefficients. In this paper, a fast feature-based algorithm is used to solve the Gaussian mixture distribution parameters. Experimental results show the validity of the model.