基于信息论分析的多尺度图像去噪方法

Multiscale image denoising method based on information-theoretic analysis

摘要:

图像多尺度统计相关模型的信息论分析表明,尺度内模型可以捕捉小波系数间的大部分相关性,较尺度间模型携带更多的信息,而通过加入父信息得到的增益则较小。为充分利用各模型提供的不同信息,提出一种基于信息融合的多尺度去噪方法,将尺度内和尺度间相关模型的优点相结合,并压制各自的缺陷。仿真结果表明,基于信息融合的方法具有更好的视觉效果和去噪性能。

Abstract:

The information-theoretic metrics evaluate the abilities of multiscale image statistical models to capture dependencies between wavelet coefficients, which indicates that intrascale model could describe most of dependencies that are more than interscale model, and the gains is small by adding parent information. To fully exploit the information provided by interscale and intrascale model, a novel multiscale image denoising method based on image fusion is proposed, which combines high denoising ability of intrascale statistical model with better performance of edge preservation of interscale model. Simulation results show that the proposed method has better visual effect and high denoising quality.