A lognormal approximation for the gray level statistics in ultrasound images

The gray level statistics of an image are important for the design of image processing algorithms. In this paper, a lognormal distribution is proposed as a practical approximation for the gray level distribution of ultrasound images. This provides a relatively simple model for describing the image, and enables using techniques originally developed for a Gaussian probability distribution. We compute various moments of the gray level distribution and show that the model predicts power laws between specific combinations of moments. These predictions, as well as two additional expressions resulting from the model, are validated using ultrasound images of various ovarian masses.