A target searching system robust on various colors using training and Biasing

In this paper, we propose a new target searching system robust on various colors in an image. The proposed system consists of two components, “Training” and “Biasing”. In “Training” component, primitive features of arbitrary objects are trained by bottom-up manner. A “Biasing” component biases feature maps with trained data as a biasing signal and finds desired targets. In the process of searching, the target bottom-up saliency and top-down saliency are both considered. We compared our system´s performance with that of previous systems in search problem, and verified that our system has robustness on various colors.