Non-linear cellular automata based edge detector for optical character images

Abstract

Design of parallel algorithms for edge detection is extremely important for image analysis and understanding. Cellular automata are the most common and simple models of parallel computation and over the last decade, numerous cellular automata techniques have already been proposed. This paper presents a novel method for edge detection of optical character images based on a variant of cellular automata, called non-linear cellular automata. The method consists of three stages and each stage is simple to understand and implement. A standard binarization technique is applied to a grayscale image and boundary conditions are added to the resultant image. Finally, non-linear cellular automata rules are designed and applied simultaneously to all pixels of the image. The suggested scheme has been validated on optical characters (handwritten as well as printed) of different languages. Furthermore, results are compared with standard edge detection techniques in terms of different performance parameters like entropy, kappa values, true positives, false negatives, etc. It is observed that the suggested scheme is superior to other standard schemes. Hence, the scheme has the potential application for character recognition.

Edge Extraction Based on Memristor Cell Neural Network With Fractional Order Template

“…The network could converge to the expected stable state according to its own dynamic characteristics in the case of reasonable design of network structure. It shows good application performance in many fields such as image denoising, edge extraction and associative memory [6]- [8].…”