Image comparison method for visual quality control based on matrix decomposition

The image processing applied on the original tile image and the replica images in serial production is used for the visual quality control. The image of a tile that contains no visual quality error represents the original image. The images of the tiles from the production line that could contain some of the visual quality errors represent the replicas. The image processing is based on the image matrix decomposition method. The matrix form of the original image is transformed to a vector. The images made by the variations of a good original image are transformed to the matrix made by such vectors. Using the SVD of the matrix of variations of the original image makes the image comparison duration shorter. The number of computer operations decreases by using SVD approximation by a lower rank matrix. The singular value decomposition calculation of the matrix is done just once so that the image processing time is shorter. Too low rank of the matrix could yield bad comparison results. The developing and testing of this method are used for an early design of the computer aided visual control.