Robustness of cellular neural networks in image deblurring and texture segmentation

摘要: Template parameters of cellular neural networks (CNNs) should be robust enough to random variability VLSI tolerances and noise. Using the CNN for image processing, one main problems is robustness a given task in real chip. It will shown that very different tasks such as 2D or 3D deconvolution texture segmentation can solved environment without significant loss efficiency accuracy under low precision (about 6–8 bits) parameters. The turns out against template noise, imperfect estimation templates parameter accuracy. The are tuned using genetic learning. These optimized depend on architecture. It was found about bits complicated multilayer deconvolution, while only 4 difficult presence noise variances. tolerance sensitivity considered implementation. Theory examples demonstrated by many results real-life microscopic images natural textures.