基于局部直觉模糊熵的偏微分图像降噪算法

Image denoising of PDE based on local intuitionistic fuzzy entropy

针对传统P-M扩散模型不能分辨梯度变化相近的平坦区和细节区的不足,提出一种新的扩散模型。该模型将局部直觉模糊熵引入扩散函数,结合梯度共同控制扩散过程,使得梯度较大的边缘区获得较小的扩散系数而保留边缘特征,对于梯度较小且相近的区域,由于细节区的局部直觉模糊熵值往往大于平坦区的局部直觉模糊熵值而获得较小扩散系数保留细节特征,弥补了传统扩散模型模糊细节特征的缺点。实验结果表明,与传统模型相比较,新模型细节信息保留的更加完整,噪声去除的更干净,视觉和量化效果均很优异。

A new model is proposed to overcome the problem in conventional P-M diffusion model that the flat areas and the detail areas with similar spatial gradient were unable to be distinguished. In the proposed model, the local intuitionistic fuzzy entropy is introduced into partial differential equation, controlling the diffusion process with the gradient together. So, the edges with bigger gradient acquire smaller diffusion coefficients and get edge-preserved, for the small gradient and similar areas, the local intuitionistie fuzzy entropies of detail areas are bigger than these of flat areas, thus these areas have small diffusion coefficients and maintain detail feature, making up for the shortcomings of traditional diffusion model that detail characteristics are always blurred. Experimental results indicate that the new algorithm can preserve more detail features and less noise than traditional models, and acquire excellent visual and quantitative effects.