摘要
为了在不降低三维块匹配算法(BM3D)效果的基础上,提高其运算速度,提出一种基于积分图的BM3D的加速算法.在新的算法中,首先利用高斯滤波器对原图像进行粗去噪,再利用积分图计算块的相似性,在块的相似度计算过程中不再进行滤波.在BM3D的第二步维纳滤波中,经过对原算法计算过程进行相应的转换,也可将积分图应用于该阶段.实验表明,改进后的算法,不但保留了三维块匹配算法在去噪方面好的性质,而且运算时间缩短了近1/4.
In order to reduce the computation time an accelerated algorithm which based on the efficient Summed Square Image (SSI) scheme is proposed in this paper. The new algorithm uses Gaussian filter to do the coarse denoising for original image first, and then SSI scheme is introduced to calculate the similarity of block. Besides, in the second process of BM3D which called wiener filtering, integral image can also be applied after the corresponding conversion of the original algorithm. Experiments have shown that the proposed algorithm not only retains the superior performance time by about 3 times of BM3D algorithm, but also reduces the operation
出处
《广西民族大学学报(自然科学版)》
CAS
2015年第2期73-80,共8页
Journal of Guangxi Minzu University :Natural Science Edition
基金
广西自然科学基金项目(2012GXNSFAA053227)
广西民族大学中国-东盟研究中心项目(KT201325)
广西民族大学研究生教育创新计划项目(gxun-chx2014088)
关键词
三维块匹配
图像去噪
高斯滤波器
积分图
快速傅里叶变换
block--matching and 3D filtering
gaussian filter
image denoising
summed square image
fast fourier transform