自适应引导滤波下图像多尺度细节增强  

Multi⁃scale detail enhancement of images under adaptive guided filtering

作　　者：李俊霖 LI Junlin(Liaoning Petrochemical University,Fushun 113001,China)

机构地区：[1]辽宁石油化工大学,辽宁抚顺113001

出　　处：《现代电子技术》2025年第9期24-27,共4页Modern Electronics Technique

摘　　要：在图像增强过程中,通常需要对图像进行去噪和平滑处理,但这一步骤往往会导致图像的边缘变得不清晰,甚至可能出现模糊的效果。因此,提出一种自适应引导滤波下图像多尺度细节增强方法。采用自适应引导滤波技术对图像实施滤波操作,在去除图像噪声和平滑图像的同时,有效保留其边缘细节特征。随后,将经过滤波处理的图像分解为两个组成部分:细节层、基础层,采用多尺度细节增强技术提升细节层图像局部细节清晰度,使用具有不同尺寸的高斯核进行卷积运算,以生成具有不同平滑效果的图像序列,从而进行多尺度细节增强。将基础层的对比度增强后与细节增强后的细节层依据一定的权重进行融合,从而生成多尺度细节增强图像。实验结果显示,所提方法能够丰富图像的细节信息,确保图像整体结构的完整性和自然过渡,使图像的色彩更加饱满;信息熵和平均梯度指标均展现出优异的表现,有力证明了该方法在图像多尺度细节增强方面的有效性。In the process of image enhancement,it is usually necessary to denoise and smooth the image,which often leads to blurring or weakening of the image edges,however.Therefore,a method for multi-scale detail enhancement of images under adaptive guided filtering is proposed.Adaptive guided filtering is adopted to filter the image,which effectively preserves the edge details of the image while denoising and smoothing the image.Subsequently,the filtered image is split into detail layer and base layer.The multi-scale detail enhancement techniques are adopted to improve the sharpness of local details of the images in detail layer,and the Gaussian kernel functions of different scales are applied for convolution operations to generate image sequences with different levels of smoothness,thus performing multi-scale detail enhancement.The base layer with enhanced contrast and the detail layer with enhanced details are fused according to a certain weight,so as to obtain multi-scale detail enhanced images.The experimental results show that the proposed method can enrich the detailed information of the image,ensure the integrity and natural transition of the overall structure of the image,and make the color of the image more full,and that the information entropy and average gradient indicators have shown excellent performance,which strongly proves the effectiveness of the proposed method in multi-scale detail enhancement of images.

关 键 词：引导滤波 自适应 多尺度 细节增强 高斯核函数 加权融合 平滑处理 

分 类 号：TN911.73-34[电子电信—通信与信息系统] TP391.41[电子电信—信息与通信工程]