Patients with prolonged stay often account for high resource consumption [1], and so, are a potential loss of revenue from a hospital management perspective. This is especially important in the case of Intensive Care Unit (ICU) and electronic ICU (eICU) patients due to their intensive need for care resources like nurses, drugs, surgeons, X-ray machines, etc.
In this post, we will introduce two broad categories of image/video quality models, and discuss two algorithms that fall under a third, hybrid, category. We will heavily borrow from concepts of Natural Scene Statics (NSS) and Information theory discussed in my earlier post on VIF.
This article is a review of existing texture characterization, identification and segmentation methods. These methods typically involve statistical methods, i.e. inferring from histograms or co-occurrence matrices, and/or signal processing methods, either simple filtering and energy-based methods, or as a method of pre-processing before applying statistical analysis.
This article is a review of a popular Image Quality Model - Visual Information Fidelity (VIF) [1] which is all based on a statistical model of natural scenes. To do this, we will first review a powerful natural scene statistics (NSS) model, the Gaussian Scale Mixture.