Quadratic Gabor filters for object detection

We present a new class of quadratic Alters that are capable of creating spherical, elliptical, hyperbolic and linear decision surfaces which result in better detection and classification capabilities than the linear decision surfaces obtained from correlation filters. Each filter comprises of a number of separately designed linear basis Alters. These Alters are non-linearly combined using an algorithm called the extended piecewise quadratic neural network (E-PQNN). In this work, we consider the use of Gabor basis Alters; the Gabor filter parameters are separately optimized; the fusion parameters to combine the Gabor Alter outputs are optimized using the conjugate gradient method; they and the non-linear combination of Alter outputs are included in our E-PQNN algorithm. We present results obtained for an infra-red (IR) vehicle detection problem.