Automatic Inspection of Defects in Plain and Texture Surfaces

The image processing used for visual inspection and quality control in a serial production is described in this paper. We used the discrete wavelet transform - DWT in our failure detection algorithm. To achieve robustness as well as good sensitivity of the algorithm, we divide the images into segments. The difference of the wavelet coefficients maxima for the given segment for images of the tile with and without defects was used for defect detection. The analysis of detection capabilities is done for different segment sizes, different detection sensitivity levels - DSL and for two orthogonal wavelets.