A visual detection method of tile surface defects based on spatial-frequency domain image enhancement and region growing

The detection of tile surface defects relies heavily on manual work and the existing automatic detection methods are difficult to be used in industrial production. In this paper, we propose a visual detection method of tile surface defects based on image enhancement and region growing algorithm. First, to eliminate the noise interference, uneven illumination and reflect light of surface during image acquisition, we propose the spatial-frequency image enhancement method. In spatial domain, the median filtering and local histogram equalization are cascaded for image denoising and contrast enhancement. In frequency domain, based on the 2D Gabor filter, the tile surface image is further processed to better eliminate the influence of uneven illumination and surface reflection. Then, we use the region growing algorithm to implement image segmentation. Based on the characteristics of tile surface defects, an automatic seed point selection method is proposed. Finally, the bidirectional integral projection algorithm is used for defect boundary detection, and based on this boundary information, the detection and marking of defect regions are realized. The detection experiments on crack, hole, pockmark and chromatic aberration defects prove the effectiveness and feasibility of the proposed method.