Robust Template Designs for Selected Objects Extraction and Masked Object CNNs with Applications

The cellular neural/nonlinear network (CNN) is a powerful tool for image and video signal processing, robotic and biological visions. The robust designs for CNN templates are important issue for the practical applications of the CNNs. The selected objects extraction (SOE) CNN and the masked object extractor (MOE) CNN are two kinds of CNNs, which have coupled templates and are able to extract or erase specific objects in processed binary images. This paper establishes two theorems for designing the robustness templates of the SOE CNNs and MOE CNNs, respectively. The two theorems provide the template parameter inequalities to determine parameter intervals for implementing the corresponding functions. Five examples are provided to illustrate the effectiveness of the methodology.