融合梯度信息的改进引导滤波

发布时间：2018-04-02 13:35

本文选题：保边平滑　切入点：引导滤波　出处：《中国图象图形学报》2016年09期

【摘要】：目的为了有效消除引导滤波平滑图像后产生的光晕现象,提出一种新型的融合梯度信息的改进引导滤波算法。方法该算法借助引导图像的梯度信息来判断图像边缘位置,并结合指数函数框架设计权值来控制不同图像区域内的平滑倍数,使改进后的引导滤波能够自适应地区分和强调边缘,从而避免边缘附近由于过度模糊所引入的光晕现象。结果与引导滤波算法相比,本文算法能在保边平滑的同时较好地抑制光晕,并在结构相似性(SSIM)评价和峰值信噪比(PSNR)评价中分别取得最高约30%和15%左右的质量提升。结论本文算法具有较好的鲁棒性,在图像平滑、图像细节增强、多曝光图像融合等多种图像处理相关应用中均有着良好的表现。
[Abstract]:Aim to effectively eliminate the halo produced by guided filtering smoothing images, a new improved guided filtering algorithm is proposed to fuse gradient information.Methods the edge position of the image is judged by the gradient information of the guided image, and the smoothing multiple in different image regions is controlled by combining the weight value of the exponential function frame.The improved guided filter can self-adaptively distinguish and emphasize the edges so as to avoid the halo caused by excessive blur near the edges.Results compared with the guided filtering algorithm, the proposed algorithm can suppress halo at the same time of edge preserving and smoothing, and achieve the highest quality improvement of about 30% and 15% in the evaluation of structural similarity and PSNR, respectively.Conclusion the proposed algorithm is robust and has good performance in many image processing applications, such as image smoothing, image detail enhancement, multi-exposure image fusion and so on.
【作者单位】：华中师范大学计算机学院;
【基金】：国家自然科学基金项目(61501198,41101425,41201413) 湖北省自然科学基金面上项目(2014CFB461) 武汉市青年科技晨光计划(2014072704011248) 华中师范大学中央高校基本科研业务费项目(CCNU14A05017)~~
【分类号】：TP391.41

【相似文献】