A method for quantification of noise non-uniformity in computed tomography images: A computational study

In computed tomography (CT), the noise is sometimes non-uniform, i.e. magnitude may vary with gradient level within image. The purpose of this study was to quantify non-uniformity in CT images using appropriate 1D and 2D computational phantoms, validate effectiveness proposed concept filtered by bilateral filter (BF), as an example a non-linear filter. We first developed Gaussian noises several levels were then added phantoms. addition, simulate real form from obtained scanner, homogeneous water phantom image used. These referred ground truth (σG). phantoms various pixel value spreads (σ) produce non-uniform noise. original (G) subtracted both noisy (IN) (IBF), magnitudes resulting for each computed. noise-gradient dependency (NGD) curve used display on It found that uniform noise, constant all gradients. However, measured dependent strength BF every large gradients decreased gradient.