Local binary pattern feature vector extraction with CNN

In this paper we present a novel approach for implementing a local binary pattern (LBP) based feature extraction system with cellular nonlinear networks (CNNs). The LBP methodology is based on transforming local binary features of an image into micro-patterns that can be used to, for example, moving object detection and face recognition and detection. We show how the LBP feature vectors can be produced using the standard CNN. Also, we show how simple modifications to the standard CNN cell can be used to make the processing of the LBPs more effective. An analog readout scheme is described and the effect of the analog readout on face recognition accuracy is simulated. The simulations are performed using the standard FERET (the facial recognition technology) database.