多尺度框架误差分散逆半调方法研究

Inverse halftoning algorithm based on multiscale framework

通过分析现有基于滤波的误差分散逆半调方法,提出了一种多尺度逆半调框架。基于该框架,引入平稳小波变换,提出了一种新的逆半调算法。首先,对半调图进行3尺度离散平稳小波变换,通过分析各尺度图像的频谱特性,构造了HIST滤波器,并结合尺度间小波系数的相关性,采用不同策略分别抑制各尺度图像中的噪声,最后利用小波反变换得到逆半调图像。实验验证了多尺度逆半调框架的正确性,并且该算法所得逆半调图像具有良好的视觉效果,能保持图像边缘,峰值信噪比也有所提高。

By studying the filtering based error diffused inverse halftoning algorithms,a multiscale framework for inverse halftoning is proposed.With the framework and stationary wavelet transform,a new inverse halftoning method is presented.Firstly,halftone images are decomposited into three scales using stationary wavelet transform.By analyzing the frequency characteristics of each scale,construcing HIST filter,and considering the correlations between inter-scales,different schemes are made to remove the noise.Experiments show that the framework is right and the proposed algorithm is able to achieve an inverse halftoning image with good visual quality,it has a better edge-preserving ability and peak signal-to-noise ratio is increased.