基于改进的HDCS的图像去噪方法

针对数字图像主要含有高斯噪声和椒盐噪声的特点,提出了一种基于改进的各向异性的混合扩散模型。传统的基于边缘增强和相干增强的模型,虽然能够有效地去除噪声;但也会存在减弱相干结构和背景的对比度等问题,同时在保持图像细节纹理方面可能会出现失真。通过在扩散方程中引入一个源项;并充分考虑它对模型中各项产生的影响,使得改进后的模型既能有效去除噪声,也能有效地保持相干结构和背景的对比度;同时在模型中引入一个偏微分方程用以获取保真项,使得图像的细节保护效果更明显。实验结果表明,该方法能达到较理想的去噪和恢复图像纹理信息的结果,而且明显改善了图像的视觉效果。

Image Denoising Method Based on Improved HDCS

According to the fact that digital images mainly contain Gauss and salt pepper noise, an anisotropic hybrid diffusion model with source term was proposed in this paper. Traditional edge-enhancing and coherence-enhancing models also had the contrast between coherent structures and the background decreased while having the image denoised. In this paper a source term was introduced into the diffusion equation, and its impacts on each term of diffusion system were fully considered. The improved model can not only perfectly denoise the noisy image, but also effectively maintain the contrast between coherent structures and the background. Experimental results demonstrated that the proposed method can achieve a desired result of denoising and restoring image texture information, and create a better visual effect.