Hue preservation based color space transformation for brightness-robust tracking

Abstract

Brightness-robust tracking is always a tough task in the applications of optical signal processing. Conventional red-green-blue (RGB) space based visual tracking always fails due to the sensitiveness to changes in brightness. Hue is a robust color feature existing in the aforementioned unstable condition, but seldom considered by researchers. In this paper, a hue preservation based color space is designed to replace RGB space and to stabilize the tracking accuracy for tracking algorithms in case of severe brightness variation. First, RGB space is mapped to hue-saturation-intensity (HSI) space and the hue component remains constant through a hue preservation principle. Then, the saturation and intensity components which are easily affected by brightness are regularized through shifting and scaling operations. When finally converted back to RGB space, connection lines that connect the three corners of the RGB cube as well as the geometric center of the lines are selected as the new color space. Experimental results demonstrate that when applied to visual tracking algorithms, our color space transformation method can improve the accuracy and stability under the condition of varying brightness, or even in the low-light-level environment.