Calculation of Optical Flow Using Color Models to Improve the Accuracy of the Identification of Objects in Motion

Currently, many applications require following objects through a sequence of images. Typically, the motion is estimated by calculating the optical flow on monochrome images because it involves fewer computational cost. However, sometimes the intensity of the pixels may not be sufficient to successfully achieve the objective of identifying moving objects. In this paper, it is proposed to use models of color, as RGB and HSV, for motion estimation method using the Lucas and Kanade method of calculation of optical flow, to improve motion estimation. As part of our contribution, it is also proposed a metric to measure the goodness of the results in order to evaluate the reliability and accuracy of optical flow calculation in each method. Finally three scenarios with different characteristics are presented to evaluate the results under different conditions.