摘要
现实生活中获取的图像可能含有不同类型的噪声,降低了图像的质量,严重影响图像应用。为了提高图像质量,需要对图像进行去噪处理。现有的偏微分方程去噪模型可以去除图像中噪声,但缺乏对纹理和边角的保护。为了有效地实现在去噪的同时保护边缘和纹理,提出了一种基于结构张量的混合阶偏微分方程去噪模型。模型利用结构张量的行列式和迹选取扩散系数,沿着结构张量的两个特征向量方向自适应扩散,既能确保图像内部纹理信息的完整性,又能减少图像边缘角点特征信息的缺失。实验结果表明:提出的模型在主观视觉和客观的评价指标上,与传统相关模型相比能更好地实现对噪声的去除和边缘细节的保护。
The acquired images contain different types of noise,which reduces the image quality and seriously affects the image application.In improve image quality,the image needs to be denoised.The existing partial differential equation denoising model can remove the noise in the image,but it cannot protect the texture and corners.To remove the noise and protect the edge and texture at the same time,a novel denoising model based on structural tensor is proposed.The proposed model uses the determinant and trace of the structure tensor of the image to determine the diffusion coefficient.The adaptive diffusion along the two feature vectors of the structure tensor can protect not only image texture information,but also image edges and corners.The experimental results demonstrate that the proposed model can remove noise while protecting detailed features.
作者
许宛如
刘奎
吴海峰
XU Wan-ru;LIU Kui;WU Hai-feng(School of Computer and Information,Anqing Normal University,Anqing 246011,Anhui,China)
出处
《合肥学院学报（综合版）》
2020年第5期26-33,共8页
Journal of Hefei University：Comprehensive ED
基金
安徽省科技重大专项“林业病虫害智能化监测预警平台研发及应用示范”(201903a06020006)
安徽省高校自然基金重点项目“基于变分PDE和非局部均值图像去噪方法”(KJ2017A353)资助
关键词
图像去噪
偏微分方程
各向异性扩散
结构张量
image denoising
partial differential equation
anisotropic diffusion
structure tenor
