一种新的复杂纹理图像的去噪方法

A New Complex Texture Images Denoising Method

针对目前复杂纹理图像去噪与保细节的矛盾,并结合以往的时域模板平滑去噪方法,提出了一种新的基于纹理分析的纹理图像去噪方法。该方法首先提出了在去噪之前应进行纹理分析,从而确定图像各个区域的纹理复杂情况;之后,又提出了使用多尺度多方向的时域模板,即根据纹理细节复杂与否来选取模板类型,并且,利用方差大小来决定细节丰富区域的线形模板方向的选取,从而增强了图像细节的保留效果。本方法运算快捷,去噪效果良好。

At present, there is a conflict between smoothing and edge preserving among complex texture images. In this paper, according to some method before, we propose an improved adaptive texture preserving denoising algorithm based on texture analysis. First it does texture analysis before smoothing, confirming the complexity of texture. Then, it uses multiscale and muhidireetional mask to deal with the issue, it needs that we chose the type of the mask basing on the texture region. And we analyses the variance to decide the direction of linear mask in the rich details region. Thus, this algorithm operates fast and easily, moreover, it is effective for both smoothing and texture preserving.