Figure 6

Segmentations of four models for four real infrared images with blurry edges and complex backgrounds, starting with a constant function${\mathbf{\psi }}_{\mathbf{0}}\mathbf{=}\mathbf{2}$. (a)-(d): original images. (e)-(h): results of the Canny algorithm (left to right: $\sigma =1,T=\left[0\text{.01,}0.3\right]$; $\sigma =1.3,T=\left[0\text{.03,}0.4\right]$; $\sigma =2,T=\left[0\text{.03,}0.6\right]$; $\sigma =1.5,T=\left[0\text{.03,}0.9\right]$). (i)-(l): results of the CV model (left to right: 500 iterations, 140 iterations, $\nu =0.015\times 255\times 255$with 500 iterations, ${\lambda }_1=1.2,\nu =0.02\times 255\times 255$ with 500 iterations). (m)-(p): results of the RSF model (left to right: 40 iterations, ${\lambda }_1=1.2,\nu =0.005\times 255\times 255$with 500 iterations, ${\lambda }_1=1.2,\nu =0.007\times 255\times 255$ with 1000 iterations, $\nu =0.06\times 255\times 255$ with 280 iterations). (q)-(t): results of our model (15 iterations). (u) and (x): 3D plots of final level set functions.