A Perception-Based Interpretation of the Kernel-Based Object Tracking

摘要: This paper investigates the advantages of using simple rules human perception in object tracking. Specifically, visual (HVP) will be used definition both target features and similarity metric to for detecting subsequent frames. Luminance contrast play a crucial role features, whereas recent advances relations between some classical concepts information theory way eye codes image metric. The use HVP well known tracking algorithm, allows us increase its efficacy following considerably reduce computational cost whole process. Some tests also show stability robustness perception-based algorithm presence other moving elements or occlusion few