Myroniv I. Information technology for recognizing text characters based on competing cellular automata

The work presents a new approach to the recognition of text characters, which is based on a new type of cellular automata (CA) - movable competing cellular automata. The mathematical model of recognition process on the basis of this type of CA has been developed; it describes their movement in the states of characters of the text and the process of competition, which allows to recognize the characters close to the image. Each character of the alphabet is represented as a system of states and has a corresponding CA with the states describing the given character. The recognition process consists of several stages. At the first stage, the scanned image is divided into lines and characters. Further, the image of characters is converted into a set of states similar to the CA. Then, cellular automata are placed in the image of each character, describing all the characters of the alphabet. At the next stage, all cellular automata begin simultaneous movement in the states of the character in accordance with their rules of transitions between neighboring states. If the transition rules do not allow a specific CA to move to the neighboring state of the character, a conclusion is made about the discrepancy between the CA and the current character, and the automaton is taken from the CA-field. As a result of this movement, cellular automata remain on the image of the character, which can realize all their states on it. There can be several automata like this, such as type C and type O cellular automaton on the O-character, type I or H on the H or I character, and type T on the T-character, etc. The next stage - the stage of competition - allows to select the cellular automaton that best describes the given character. The competition "wins" the CA, that has the most of realized states on the character. Reading the unique token of such CA, the system gives out the user a recognition result. In case when the competition does not reveal a definite result, the user is given an opportunity to make an independent choice. The mathematical model of the behavior of competing cellular automata is developed. The information technology of character recognition based on competing cellular automata is created.