Fig. 1. Illustration of frequency domain segmentation for curvelet transformation

Fig. 2. Simulation results of weak texture signal images of submarine's V-wake in direct course. (a) Weak texture image in Ref. [7]; (b) weak texture image in Ref. [8]; (c) weak texture image in Ref. [15]; (d) weak texture image in Ref. [16]; (e) weak texture image in Ref. [7] with background; (f) weak texture image in Ref. [8] with background; (g) weak texture image in Ref. [15] with background; (h) weak texture image in Ref. [16] with background

Fig. 3. Contrast and average of grayscale of V-shape wake

Fig. 4. Directional component screening for V-shape wake. (a) Directional component sizer in frequency domain; (b) weak texture after directional component screening

Fig. 5. Optimized threshold screening for V-shaped wake

Fig. 6. Edge extraction of gradient operator for V-shaped wake

Fig. 7. Flowchart of algorithm

Fig. 8. Different V-shaped wakes and edge extraction results of different algorithms. (a) Model in Ref. [7] added with background; (b) result of proposed algorithm for Fig. 8(a); (c) result of curvelet transform for Fig. 8(a); (d) result of wavelet transform for Fig. 8(a); (e) model in Ref. [8] added with background; (f) result of proposed algorithm for Fig. 8(e); (g) result of curvelet transform for Fig. 8(e); (h) result of wavelet transform for Fig. 8(e); (i) model in Ref. [15] added with background; (j) result of proposed algorithm for Fig. 8(i); (k) result of curvelet transform for Fig. 8(i); (l) result of wavelet transform for Fig. 8(i); (m) model in Ref. [16] added with background; (n) result of proposed algorithm for Fig. 8(m); (o) result of curvelet transform for Fig. 8(m);(p) result of wavelet transform for Fig. 8(m)

Fig. 9. Algorithm adaptability of texture contrast

Fig. 10. Algorithm adaptability of a priori frequency screening direction