An eight layer cellular neural network for spatio‐temporal image filtering

Abstract

SUMMARYSpatio-temporal ÿlters are critical components of biologically inspired or neuromorphic algorithms for image motion analysis. In this paper, we describe eight layer cellular neural network architectures that can be used to implement these ÿlters. Despite the apparently large number of layers, we describe how these architectures can be implemented e ciently using weak inversion transistor circuits. Integrating both spatial and temporal ÿltering into a single network reduces hardware complexity in comparison with an architecture that cascades separate spatial and temporal ÿltering stages. In addition, by considering spatial and temporal ÿltering jointly, we can obtain ÿlters with enhanced velocity selectivity, as well as more robust population responses to moving image input.

Analysis of the interaction between the retinal ON and OFF channels using CNN‐UM models

“…If we recombine the rectified signals before the blurring (Figure 8), then the bipolar cells' original biphasic signal is reconstructed ( Figure 6). For the ON system this can be seen in Equation (4) for the OFF in Equation (5). …”

On Analog Networks and Mixed-Domain Spatio-Temporal Frequency Response

“…The frequency response of (18) is given by (19) note that we only consider real coefficients in (10). Let (20) and (21) Both and incorporate the factor . Depending on the value of the integer , one of the terms will be real and the other imaginary.…”

“…8. Despite its rich dynamics for system modeling [19], limited research has been carried out in exploiting it for linear spatio-temporal filtering until recently [20]. To contribute to the revelation of the spatio-temporal filtering capability of MLCNNs in the context of STTFs, we consider a two layer linearized MLCNN as an example and distinguish between two cases: 1) with feedforward interlayer connections only and 2) with both feedforward and feedback interlayer connections.…”

Modular Cellular Neural Network Structure for Wave-Computing-Based Image Processing

“…There are, however, a few studies based on the multilayer CNN architecture. They showed the usefulness of their structure for center point detection and skeletonization and discussed its stability [34]. Balya and others proposed a three-layer CNN structure for modeling the mammalian retina model and discussed the stability of their structure [25].…”