摘要
针对图像去噪过程存在边缘保持与噪声抑制的矛盾,提出一种改进的偏微分方程图像降噪算法。将差分曲率引入各项异性扩散方程中,构造出新的扩散系数函数,其可以较好区分图像的边缘、平坦区域以及噪声,使算法在去除噪声的同时,保留更多弱边缘和纹理等细节信息。实验结果表明,该算法与其它基于偏微分方程的图像降噪算法相比,在客观的质量评价方面,提高了图像的信噪比和峰值信噪比;在主观视觉效果方面,在滤除噪声的同时保留了更多的弱边缘等细节信息,该算法实现了较好的图像降噪效果。
Concerning the contradiction between edge-preserving and noise-suppressing in the process of image denoising,an improved partial differential equation（PDE）for image denoising was proposed.Difference curvature was combined with anisotropic diffusion equation,and a new diffusion coefficient function was constructed.As the new diffusion coefficient function effectively distinguished edges from ramp regions and isolated noise,the model preserved more details such as weak edges and textures while getting rid of the noise.Experimental results show that,compared with other image denoising methods based on PDE,the proposed method improves signal-to-noise ratio（SNR）and peak-signal-to-noise ratio（PSNR）in the terms of objective quality evaluation,and preserves more detail information such as weak edges while removing the noise on the aspect of subjective visual effect.Therefore,the presented algorithm obtains better image denoising results.
出处
《计算机工程与设计》
北大核心
2015年第7期1850-1854,共5页
Computer Engineering and Design
基金
国家自然科学基金项目(61071192
61271357
61171178)
山西省国际合作基金项目(2013081035)
山西省研究生优秀创新基金项目(2009011020-2
20123098)
中北大学第十届研究生科技基金项目(20131035)
关键词
图像去噪
偏微分方程
各项异性扩散
差分曲率
弱边缘
image denoising
partial differential equation
anisotropic diffusion
difference curvature
weak edges
