基于局部梯度统计的自适应各向异性扩散  

Adaptive anisotropic diffusion based on local gradient statistics

作　　者：徐岩松 张爱军[1] 杨萌 XU Yansong;ZHANG Aijun;YANG Meng(School of Mechanical Engineering,Nanjing University of Science and Technology,Nanjing 210094,China)

机构地区：[1]南京理工大学机械工程学院,南京210094

出　　处：《激光杂志》2023年第8期60-64,共5页Laser Journal

基　　金：江苏省研究生科研与实践创新计划项目(No.SJCX22_0103)。

摘　　要：针对现有各向异性扩散模型在图像去噪中存在的纹理和边缘细节丢失问题,提出一种基于局部梯度统计的自适应各向异性扩散模型。本模型利用邻域像素梯度的统计信息适应的调整中心像素扩散函数,以更准确地针对噪声和边缘执行差异性操作;扩散函数的生成仅发生在迭代开始前,保证了较小的总计算量。实验结果表明,与两种现有对照模型相比,本模型的输出图像在充分去噪的同时可以保留更多的原始图像细节,峰值信噪比提高约2.24 dB,结构相似度提高约11%,而耗时仅为对照模型的30%~40%,在改善去噪质量的同时取得了较高的处理效率。An adaptive anisotropic diffusion model based on local gradient statistics is proposed to solve the prob-lem of texture and edge detail loss in existing anisotropic diffusion models in image denoising.This model adaptively adjusts the central pixel diffusion function by using the statistical information of the neighborhood pixel gradient to per-form more accurate differential operations against noise and edges.The generation of diffusion function only takes place before the start of iteration,which ensures a small amount of total calculation.The experimental results show that the proposed model improves the quality of image denoising and achieves high processing efficiency.Compared with the two contrast models,this model can retain more original image details while fully denoising,and achieve faster processing speed.

关 键 词：各向异性扩散 PM模型 图像去噪 保边滤波 

分 类 号：TN209[电子电信—物理电子学]