Fig. 1. Flow chart of defect measurement algorithm based on spectral estimation and multispectral technique

Fig. 2. Defect measurement system. (a) Schematic of defect measurement system; (b) picture of experimental system

Fig. 3. Defect image of selected area in the same field of view

Fig. 4. Number of defects of single spectral image in the same area at different wavelengths

Fig. 5. Number of defects of single spectral image in the same area under different wavelengths and de-noising intensities

Fig. 6. Selected defects 1, 2, and 3 in measurement image

Fig. 7. Pixel areas of defects 1, 2, and 3 in single spectral image at different wavelengths

Fig. 8. Defect detection results of single spectral image at different wavelengths.(a) Original image; (b) λ=440 nm; (c) λ=500 nm; (d) λ=505 nm

Fig. 9. Number of defects of multispectral images with different wavelength combinations. (a) λ=475 nm; (b) λ=500 nm; (c) λ=565 nm; (d) λ=650 nm

Fig. 10. Pixel areas of defects 1, 2, and 3 of multispectral images with different wavelength combinations.(a) λ=475 nm; (b) λ=500 nm; (c) λ=565 nm; (d) λ=650 nm

Fig. 11. Defects 1, 2, and 3 used for small size recognition in multispectral images with different wavelength combinations. (a) Selected defects; (b) detected wavelengths of defects 1 and 2; (c) detected wavelengths of defects 1 and 3; (d) detected wavelengths of defects 2 and 3