A novel and automatic character extraction and recognition for Sudoku puzzle solving

Abstract

This paper proposes a novel algorithm to automatically solve a Sudoku puzzle on images taken from magazines or computer game software. The process involves preprocessing the image, extracting characters, recognizing characters and solving puzzle, which all are specifically designed for Sudoku. The algorithm effectiveness is verified with images taken from cameras. In this paper, a pixel to pixel comparison or a three layer Neural Network is used to recognize digits from 1 to 9 in Sudoku puzzles.