A unified framework for local visual descriptors evaluation

Local descriptors are the ground layer of recognition feature based systems for still images and video. We propose a new framework to explain local descriptors. This framework is based on the descriptors decomposition in three levels: primitive extraction, primitive coding and code aggregation. With this framework, we are able to explain most of the popular descriptors in the literature such as HOG, HOF, SURF. We propose two new projection methods based on approximation with oscillating functions basis (sinus and Legendre polynomials). Using our framework, we are able to extend usual descriptors by changing the code aggregation or adding new primitive coding method. The experiments are carried out on images (VOC 2007) and videos datasets (KTH, Hollywood2 and UCF11), and achieve equal or better performances than the literature.