An Approach to Solve Sudoku Using Image Processing and a Genetic Algorithm

Abstract

Sudoku is logic-based puzzle and quite a popular game. The 9 by 9 square grid Sudoku puzzle is the one that we have taken into an account. For solving sudoku puzzle, we have used image processing along with genetic algorithm. In the first phase, the raw and unsolved image is first gray scaled and then Gaussian blur is applied to reduce noise. Then it goes through a main frame extraction stage where adaptive thresholding, maximum contour detection, perspective transformation and small grid segmentation are performed respectively. The individual 81 grids are passed to the CNN model for digit recognition. Then finally, genetic algorithm is used to find the optimal solution of Sudoku problem.