基于多物理场的图像处理中的点扩散函数

Point Diffusion Function in Image Processing Based on Multi-physical Fields

P-M模型是经典的图像处理模型,其扩散过程中的"传导系数"依赖于图像的局部特性--梯度模值。由于噪声的影响,图像的局部特性很难被准确的估计,通常采用高斯滤波对图像进行预处理。为了避免由梯度和Gaussian预滤波造成的问题,提出一种多物理场图像处理模型。该模型赋予了图像一定的物理性质,利用图像的局部物理特性构造扩散函数,并对扩散函数的特征进行了分析。实验结果表明,该扩散函数具有较强的保边能力和较快的收敛速度,当峰值信噪比(PSNR)和边缘保持指数(EPI)与P-M相当时,所用时间大大缩短,能更好地去除噪声并保持图像细节和边缘信息。该模型是一种自适应扩散过程,避免了滤波前的先验估计。

The P-M model is a classical image processing model,diffusion process of which depends on the local characteristics of the image.Due to the influence of noise,Gaussian filter is usually used to preprocess the image.A multi-physical field image processing model is proposed in this paper,which endows the image with some physical properties.The diffusion function is constructed by using the local physical characteristics of the image, and the characteristics of the diffusion function are analyzed.Experimental results show that the diffusion function has strong edges preserving ability and fast convergence speed, and the time used is greatly reduced when the peak-signal-noise-rate (PSNR) and edge preserve index (EPI) are comparable to P-M.The model is an adaptive diffusion process to avoid the prior estimation before filtering.