According to the properties of noise and the features of grayscale variation in dual - energy X - ray liquid images , the paper proposed a method combining improved adaptive median filtering and NL - means filtering with noise statistic estimator to cancel both Gaussian noise and pulse noise. Firstly, image pixels are classified into impulse noisy pixels、edge pixels and Gaussian noisy pixels by noise detection operators. The paper a-dopted improved adaptive median filtering to process impulse noisy pixels and edge pixels, NL - means filtering with noise statistic estimator to process Gaussian noisy pixels. Then, a method of image quality assessment based on material feature curves is applied, the results are compared with the calibration curve. Extensive simulation experiments demonstrate that new algorithm can preserve edge detail, material features and save time. The method of image quality assessment combines objective and subjective factors to meet needs of practical applications.%针对双能X射线液体图像灰度连续变化的特性和主要受高斯噪声、脉冲噪声干扰的特点,提出一种改进的自适应中值与非局部均值结合的滤波方法.首先,利用改进自适应中值方法滤除噪声检测出的边缘、脉冲噪声污染区的噪声；再利用带有噪声估计的非局部均值方法(NL-means)平滑高斯噪声.提出了基于材料特征曲线的滤波器性能评价方法,滤波结果与标准曲线进行对比.实验结果表明,提出的滤波器保护了边缘、材料信息,提高了速度；滤波器评价方法考虑主观与客观因素,满足了应用需要.
展开▼
