高级检索

基于半高斯梯度算子的尺度自适应纹理滤波

Scale Adaptive Texture Filtering Based on Semi-Gaussian Gradient Operator

  • 摘要: 为了解决现有纹理平滑方法不能在抑制纹理的同时保持细窄结构的问题,提出一种基于半高斯梯度算子的尺度自适应纹理滤波算法.首先提出一种具有结构保持和纹理抑制效果的半高斯梯度算子,其幅值能够较好地区分结构和纹理,方向能够较好地反映结构法向;然后基于半高斯梯度的幅值为每个像素估计最优的滤波尺度,并借助高斯滤波得到一幅大部分纹理得到抑制的引导图像;最后在双边滤波的基础上引入基于半高斯梯度方向的滤波方向核,提出一种基于引导图像的三边纹理滤波算法,能够有效地避免图像平滑结果中存在的结构附近毛刺残留现象.大量实验结果证明,该算法可以在抑制纹理的同时保持结构的完整性,尤其是在保持细窄结构以及结构周围的整洁等方面超越了已有的方法.

     

    Abstract: In order to solve the problem that the existed texture smoothing methods can not maintain narrow structure while removing texture,a scale adaptive texture filtering algorithm based on semi-Gaussian gradient operator is proposed.Firstly,we design a semi-Gaussian gradient operator with texture suppressing and structure preserving characteristics.Its amplitude can distinguish structure from texture,and its direction represents the normal direction of the structure.Secondly,the optimal smoothing scale for each pixel is estimated based on the amplitude of the semi-Gaussian gradient,and the Gaussian filter is utilized to obtain a guidance image in which most of the texture are diminished.Finally,a filtering direction kernel based on the semi-Gaussian gradient direction is introduced to the bilateral filter,and a trilateral texture filtering algorithm based on the guidance image is put forward,which can effectively avoid the burr near the structure.Experimental results show that our algorithm can guarantee the structural integrity while flattening the texture,and show superior performance in maintaining the narrow structures and keeping the structures clean.

     

/

返回文章
返回