基于梯度直方图的图像自适应滤波去噪  

Adaptive Filtering Method for Image Denoising Based on Gradient Histogram

作　　者：焦莉娟[1] 王文剑[2] 裴春琴[1] JIAO Lijuan;WANG Wenjian;PEI Chunqin(Computer Department,Xinzhou Teachers University,Xinzhou 034000,China;School of Computer and Information Technology,Shanxi University,Taiyuan 030006,China)

机构地区：[1]忻州师范学院计算机系,山西忻州034000 [2]山西大学计算机与信息技术学院,山西太原030006

出　　处：《山西大学学报（自然科学版）》2023年第1期141-146,共6页Journal of Shanxi University(Natural Science Edition)

基　　金：国家自然科学基金(62076154,61673249);山西省国际科技合作重点研发计划项目(201903D421050);山西省科学社会哲学规划项目(2021YY214);忻州师范学院科研项目(WTSYJ202101)。

摘　　要：为了解决自适应滤波器在图像去噪中因需要噪声检测以及人工设置阈值,从而影响去噪效果的问题,提出基于梯度直方图的自适应滤波方法。首先,对噪声图像均值滤波后的初始去噪图进行计算,得到梯度直方图。然后,通过对梯度直方图曲线形状进行分析,计算出分割性最优的点作为阈值。最后,用计算得到的阈值与图像信息的局部变化率相结合,建立尺度自适应调节的滤波模板,对噪声图像进行滤波去噪。实验结果表明,本文算法针对不同噪声类型和不同强度的含噪图像去噪效果均有提升,并且可与其他算法相融合,对自适应类算法的改进具有普适性价值。Noise detection and manual setting threshold may lead to poor denoising effect in the process of image denoising using adaptive filters. Therefore, an improved algorithm is proposed. Firstly, a gradient histogram is calculated from the initial denoising image obtained by mean filtering the noisy image. Secondly, two points with the best segmentation are figured out as thresholds by analyzing the gradient histogram curve. Finally, images are denoised using the adaptive filtering template, which is created by combining calculated thresholds above with local variance ratio of these images. The experimental results show that the proposed algorithm can improve the denoising effect of noisy image with different noise intensity, and it can be integrated with other ones to improve the denoising effect for different noise types and intensities. The proposed method has universal value for the improvement of adaptive algorithms.

关 键 词：图像去噪 梯度直方图 自适应滤波器 阈值设置 

分 类 号：TP394.1[自动化与计算机技术—计算机应用技术]