一种基于前后向扩散的图像去噪与增强方法

基金项目: 总装预研基金资助项目(51322020703)

详细信息

A Forward and Backward Diffusion Based on Image Denoising and Enhancement Method

¹School of Electronics and Information Engineering,Harbin Institute of Technology,92 West Dazhi Street,Harbin 150001,China

Funds: 总装预研基金资助项目(51322020703)

摘要: 将各向异性前向扩散和后向扩散进行深度融合,提出了一种基于前后向扩散的图像去噪与增强方法。该方法利用非线性结构张量代替直接的梯度估计,增强对噪声的鲁棒性;将沿梯度方向增强的冲击滤波项改为按照自适应设定的阈值,进行前向扩散去噪和后向扩散增强的相互转换;对于角型结构,在与梯度垂直方向同样进行后向扩散,以增强角型纹理。实验结果表明,本文算法不仅有效去除了噪声,而且增强了纹理。
- 各向异性扩散 /
- 冲击滤波器 /
- 结构张量 /
- 去噪
Abstract: A novel forward and backward diffusion based image denoising and enhancement method is proposed.In the proposed method,the direct gradient computation is replaced by a nonlinear structure tensor(NLST) to increase the robustness to noise;The shock filter item along the gradient direction is replaced by a forward or backward diffusion according with the adaptive threshold value of gradient;The corner structure is also enhanced by performing a backward diffusion in corner region.Two experimental results show that the proposed algorithm not only removes noise,but also efficiently enhances image texture details and no artifacts are produced.
- anisotropic diffusion /
- shock filter /
- structure tensor /
- denoising