User:JonasEdi

In computer vision field, statistical color representation models (SCM) are used for vision and image processing tasks, such as tracking, segmentation, and object recognition by driving the active contour models, also known as snakes, to delineate the object outline. Statistical model is constructed by performing cluster analysis of seed regions in image or image sequence. Given the RGB value vector, SCM provides "goodness" value, which shows likelihood if a pixel belongs to the region being modeled.

Trivariate Gaussian Model
The probability density function for multivariate Gaussian is given by:
 * $$N(\vec x, \bar x, S) = \frac {1}{\sqrt{(2\pi)^n \left | S \right \vert}} \exp {\left [ -\frac {1}{2} (\vec x - \bar x)^T S^{-1}(\vec x - \bar x)\right],}$$

where $$ \vec x$$ is 3-dimensional measurement vector, lying in RGB colour cube, $$\bar x$$ is the mean of the distribution, n is the dimension of the data being modelled, S is the covariance matrix and $$\left | S \right \vert$$ is its determinant.

Chromaticity Models

 * Zhu and Yuille's SCM uses planar representation of chromaticity aligned with the mean of the sample distribution, and projects RGB data orthogonaly, making the model cylindrical in the RGB color cube.
 * Healey's model derived by normalizing parameters of a trivariate Gaussian model of the distribution of RGB vectors.
 * Bingham distribution SCM is a normal distribution on the surface of the unit sphere in 3-dimensional space. It ignores specular components of the reflected light.

Applications
Area tracking can be widely used to help autonomous systems to percept the environment. It can be used to recognize roads, path markers, moving obstacles, etc.Source describes SCM application for skin detection to provide adult material filter.