Fig. 1. Morphologies of point targets with (a) one pixel, (b) two pixels, (c) three pixels, (d) four pixels, and (e) multiple pixels

Fig. 2. (a) Complex cloud background; (b) enlarged view of cloud background edge; (c) three-dimensional grey-scale map corresponding to Fig. 2(b)

Fig. 3. Structural elements in eight-directions. (a) 0° ; (b) 45°; (c) 90°; (d) 135°; (e) 180°; (f) 225°; (g) 270°; (h) 315°

Fig. 4. Candidate point configuration detected by 0°-direction structural element

Fig. 5. Comparison among detection results. (a) Original infrared image; (b) detection result by TH transformation; (c) detection result by omnidirectional morphology

Fig. 6. Sketch map of four directional vectors of candidate points

Fig. 7. Schematic of cross-pixel point target

Fig. 8. Gray level images of (a) point target imaging at pixel center, (b) point target across 4 pixels, and (c) noisy point

Fig. 9. Energy concentration degree of point targets

Fig. 10. Image acquisition equipment

Fig. 11. Target detection results. (a)-(c) Original infrared images; (d)-(f) results after adaptive threshold detection; (g)-(i) results after removal of background edges; (j)-(l) results after removal of noise

Fig. 12. Signal-to-noise ratio of point targets

Fig. 13. Processing results from different algorithms. (a) Max-median filter algorithm; (b) DoG scale-space detection algorithm; (c) BM3D algorithm; (d) GMM algorithm