基于低阶图像块和迭代的射线检测图像噪声水平估计

Estimating the Noise Levels in Radiographic Images Based on Low-Rank Image Patches and Iterations

针对射线检测图像噪声水平估计时过分依赖混合噪声模型和估计易失效的问题,提出一种基于低阶图像块和迭代的射线检测图像噪声水平估计方法.首先,根据射线检测图像中存在胶片颗粒度噪声、暗电流噪声和散粒噪声等的特点,对射线检测图像的噪声水平进行初始估计;然后,通过低阶图像块的选择和纹理强度阈值的计算,克服噪声水平估计时容易存在的过估计和欠估计问题;最后,通过多次噪声水平估计的迭代,精确估计射线检测图像的噪声水平(即噪声标准差).为了验证所提方法的性能,将它与常用的基于尺度不变和基于局部方差和偏态的噪声水平估计方法进行了对比.实验结果表明,所提噪声水平估计方法在对射线检测图像噪声水平估计时具有较好的精度和效率.

Aiming at the problems of excessively depending on hybrid noise model and easily estimation failure when estimating noise levels in radiographic images,a noise level estimation method based on low-rank image patches and iterations was proposed.Firstly,according to the characteristics of film granularity noise,dark current noise and loose particle noise in an radiographic image,the noise level in an radiographic image was primarily estimated.Secondly,the low-rank image patches were selected and the texture strength threshold was calculated to overcome the overestimate and underestimate problems.Finally,the noise level in the radiographic image was accurately estimated by multiple estimation iterations.In order to verify the performance of the proposed method,some comparative experiments had been performed with the common noise level estimation methods based on scale invariance and local variance and skewness.The experimental results reveal that the proposed noise level estimation method achieves higher accuracy and efficiency when estimating the noise levels in radiographic images.