一种改进的双变量收缩模型图像去噪

An improved image denoising method for bivariate shrinkage model

针对噪声图像低频子带含有噪声的特点,给出了一种改进的局部自适应双变量收缩模型的图像去噪算法,对于高频子带用局部自适应双变量模型进行去噪,而对低频子带用具有局部自适应的高斯模型进行去噪。该算法既体现了尺度内的聚类性,又体现了尺度间的相关性且具有很好的局部自适应性,在实验中用离散小波变换进行去噪。实验结果表明,这种改进的算法无论从峰值信噪比,还是从主观视觉效果上都要优于传统的去噪算法。

An improved image denoising method for local adaptive bivariate shrinkage model is proposed in this paper ac-cording to the characteristics that the low-frequency subband of noise image contains noise. The high-frequency subband is de-noised by locally adaptive bivariate shrinkage model,and the residual low-frequency subband is denoised by locally adaptive Gaussian model. This method can reflect both the clustering performance of intra-scale and the correlation of inter-scale,and has good local adaptive property. The discrete wavelet transform was used to denoise in a experiment. The experimental results show the improved algorithm is more superior to the classical methods in both PSNR and subjective visual effect.