基于自适应中值滤波半径的纹理滤波算法  

Texture Filtering Based on Adaptive Median Filtering Radius

作　　者：郭和炀 刘俊如 赵汉理[1] 

机构地区：[1]温州大学智能信息系统研究所,温州325035

出　　处：《现代计算机》2017年第24期15-19,共5页Modern Computer

基　　金：温州大学研究生创新基金项目(No.16-29)

摘　　要：图像纹理滤波在保持图像中的显著结构信息的同时有效地过滤图像中的纹理信息。传统的纹理滤波方法很难平衡好复杂纹理过滤能力和结构边界特征保持能力之间的效果。提出一种改进地基于方向的相对总变差模型来描述图像中的纹理特征尺度,可以有效地区分纹理特征和结构边界。基于最大-最小图像栈方法计算出图像中各个像素的自适应中值滤波半径,并运用自适应中值滤波生成一幅引导图像。基于引导图像对输入图像做联合双边滤波操作获得结构保持的纹理滤波图像。实验对比数据证明该算法既能有效地平滑纹理细节信息,也能有效地保持显著结构特征。Image texture filtering can preserve salient structure information and effectively filter textural information in the image. Traditional texture filtering algorithms are difficult to balance effects between the capability of filtering complex textures and the capability of protecting fea- tures in structural boundaries. Proposes an improved direction-based relative total variation model to measure textural features and to effec- tively distinguish texture features and structural boundaries. By using the maximum-minimum image stack, the adaptive median filtering radius for each pixel is computed and then the adaptive median filtering is employed to generate a guidance map. The joint bilateral filter- ing is applied to the input image to produce the structure-preserving texture filtered image. Experimental comparisons and results validate that the proposed algorithm can not only smooth textural details but also protect salient structure features.

关 键 词：纹理滤波 中值滤波 结构保持 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]