Extended LBP Operator to Characterize Event-Address Representation Connectivity

PABLO NEGRI

Lima

Congreso; Iberoamerican Congress on Pattern Recognition; 2016

Pontífica Universidad Católica del Perú

Address-Event Representation is a flowering technology that can change the visual perception of the computer vision world. This paper proposes a methodology to associate the input data from this kind of sensors. A new descriptor computed using an extended LBP operator seeks to characterize the connectivity of the asynchronous incoming events in a two dimensional space. Those features can be organized on histograms and combined with others descriptors, as histograms of oriented events. They can be the input of traditional classifiers to detect or recognize objects from the scene.