Improvement of the Segmentation by means of the Addition of a Color Vector and the Modeling of the Dispersions of Saturation and Hue

In this paper, an alternative that allows to improve the color image segmentation in the HS sub-space (HSI space) is presented. The authors propose to add a color vector in the RGB space with the objective of increasing the separation between classes in the HS plane. This vector is considered optimum and the chromatic C₁C₂ sub-space (YC₁C₂ space) is used to define it. In order to minimize the variances computing times, the authors also propose to estimate the dispersions of the classes projected in the HS plane from their dispersions in the C₁C₂ linear chromatic plane, using the geometric-analytical relationships between both spaces. The proposal presented in this paper has been designed to be applied in real-time on each frame of a color image sequence. Its effectiveness has been tested in applications where a reduced contrast between the background color and the color of the object to segment exists, and when the size of the object to segment is very small in comparison with the size of the captured scene.