摘要
小波域隐马尔可夫树(HMT)模型被广泛应用于统计信号和图像处理中,它成功描述了真实图像小波系数在尺度之间的相关性和依赖性,很好地体现了小波变换的延续性和非高斯性。这里通过构建图像小波域HMT模型,在应用期望最大(EM)算法估计HMT模型的参数之后,对小波系数进行贝叶斯估计达到去除噪声的目的。实验结果表明,去噪效果好于其他小波去噪算法。
Wavelet-Domain HMT models has been applied in statistic signal and image processing. The features of the wavelet coefficients between the scales of real-world image are captured, such as correlation and persist- ency and it is incarnated the Persistency and NonGaussianity properties of the wavelet coefficients. The Wavelet-Domain HMT model was built and the EM algorithm was used to estimate the parameters of the HMT model. At last Bayesian MAP was used to estimate the wavelet coefficients. The experimental results showed that the denoising method was more efficient than the other method based on wavelet transforms.
出处
《测绘科学技术学报》
北大核心
2010年第2期120-122,共3页
Journal of Geomatics Science and Technology
关键词
小波变换
隐马尔可夫树模型
贝叶斯估计
图像滤波
四叉树
wavelet transforms
HMT (Hidden Markov Tree)
Bayesian estimation
image denoising
Quadtrees