作者: T. Serrano-Gotarredona , B. Linares-Barranco , A.G. Andreou
DOI: 10.1109/IJCNN.2000.860750
关键词: Computational science 、 EPROM 、 Machine vision 、 Composite image filter 、 Convolution 、 Very-large-scale integration 、 Filter (signal processing) 、 Computer hardware 、 Kernel (image processing) 、 Real time vision 、 Chip 、 Computer science
摘要: A neural architecture that implements a programmable 2D image filter has been presented. The allows to implement any F(p,q) decomposable into x-axis and y-axis components = H(p)V(q) such the product can be approximated by signed minimum. Positive negative values of H(p) V(q) programmed. requires an address even representation (AER) input. This rotate convolution kernel angle. Circuit simulation results critical were given. System-level behavioral simulations 128x128 array have included which validate proposed approach.