21 April 2003 On-chip training for cellular neural networks using iterative annealing
Author Affiliations +
Cellular Neural Network-Universal Machines (CNN-UM) are analog devices, which are excellently suited for image processing. A big challenge thereby is the determination of CNN templates for special image processing tasks. In many cases appropriate templates can only be found by a parameter optimization. The determination of templates for complex applications in the area of CNN is usually performed by using a CNN software simulator. Unfortunately, in many cases the determined templates cannot be used in hardware realizations of CNN caused by realization effects. In order to find robust templates, which are not only working on CNN simulators, but also on hardware implementations, we present in this contribution a new kind of on-chip-multi-template-training. Furthermore, as a possible application, we will also present a CNN-based solution of the problem of Pattern Matching, which is a processing step in many areas of image processing, like e.g. in Motion Estimation, Image- and Video-Compression.
© (2003) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Dirk Feiden, Dirk Feiden, Ronald Tetzlaff, Ronald Tetzlaff, } "On-chip training for cellular neural networks using iterative annealing", Proc. SPIE 5117, VLSI Circuits and Systems, (21 April 2003); doi: 10.1117/12.498954; https://doi.org/10.1117/12.498954

Back to Top