Fig. 1. Image of ceramic tile. (a) Crack image; (b) surface gray distribution map

Fig. 2. Process of surface crack detection of ceramic tile based on PCA method. (a) Red channel image of ceramic tile; (b) results of sample centralization; (c) covariance matrix of centralization matrix displays with two-dimensional image; (d) projection matrix displays with two-dimensional image; (e) dimension-reduced post-sample matrix displays with two-dimensional image; (f) reconstruction of h=20; (g) results of difference between graph (f) and graph (a); (h) results of crack detection

Fig. 3. Effect of principal component h on defects. (a) Red channel image of ceramic tile; (b) h=10 reconstruction image; (c) h=15 reconstruction image; (d) h=20 reconstructed image; (e) h=25 reconstructed image; (f) h=30 reconstructed image

Fig. 4. Crack detection process of ceramic tile in texture area. (a) Red channel image of ceramic tile; (b) h=20 reconstruction image; (c) result of difference between graph (b) and graph (a); (d) result of crack detection

Fig. 5. Test results of different algorithms. (a) Red channel image of ceramic tile; (b) Canny operator; (c) discrete wavelet transform; (d) automatic area growth; (e) our algorithm

Table 1. Detection rate of different algorithms