Fig. 1. Image of ceramic tile.(a) Typical ceramic tile crack image; (b) 3D image of surface gray distribution of a typical ceramic tile

Fig. 2. Multi-scale Hessian matrix filtering ceramic tile surface crack detection process

Fig. 3. Image multi-scale filtering. (a) Original image of crack area I^σ(σ=0); (b) spatial scale image I^σ(σ=1); (c) spatial scale image I^σ(σ=5); (d) spatial scale image I^σ(σ=10); (e) spatial scale image I^σ(σ=15); (f) spatial scale range image I^σ(σ=1)~I^σ(σ=15)

Fig. 4. Test results of crack area defect. (a) Hessian crack enhancement image; (b) crack binarization image

Fig. 5. Detection process of ceramic tile crack. (a) Red channel image of ceramic tile; (b) spatial scale range image I^σ(σ=1)~I^σ(σ=15); (c) Hessian crack enhancement image; (d) result of crack detection

Fig. 6. Test results of different algorithms. (a) Red channel image of ceramic tile; (b) method of Ref. [8]; (c) method of Ref. [9]; (d) method of Ref. [11]; (e) method of Ref. [21]; (f) method of Ref. [22]; (e) proposed method