Fig. 1. Diagram of system light source and camera arrangement. (a) Sectional view; (b) diagram of right LED array

Fig. 2. Irradiance distribution image corresponding to the best illumination parameters

Fig. 3. Comparison of non-uniform illumination and uniform illumination. (a) Circular aggregate LED array (with light diffusion plate); (b) symmetrical two-board LED array; (c) shooting effect of circular array; (d) shooting effect of two-board array

Fig. 4. Flow chart of vascular 3D imaging algorithm

Fig. 5. Intermediate result images of monocular image vascular skeleton extraction algorithm. (a) Anterior region of neck; (b) near the sternoclavicular joints; (c) forearm; (d) back of hand

Fig. 6. Concept of pixel connectivity. (a) Pixel neighborhood; (b) ambiguities of 8-connected regions; (c) redundancy connecting points and fork points

Fig. 7. Physical diagram of convergent binocular near-infrared imaging system

Fig. 8. Phantom model. (a) Vascular model; (b) vascular model covered by muscle tissue; (c) covering skin

Fig. 9. Extraction results. (a) Stereo rectified result; (b) CLAHE result; (c) Frangi result; (d) binarization result; (e) refinement result; (f) skeleton branch pruning result

Fig. 10. Overlay effect of skeleton extraction result of phantom model and stereo rectification image

Fig. 11. 3D reconstruction result of vascular skeleton of phantom model

Fig. 12. Vascular skeleton extraction and 3D reconstruction results. (a) ROI region; (b) skeleton extraction results; (c) 3D reconstruction results

Fig. 13. Local amplification drawing of hand 3D reconstruction results

Fig. 14. Comparison of elapsed time between GPU and CPU algorithms in the vascular skeleton extraction

Fig. 15. Stereo matching algorithm evaluation. (a) Elapsed time; (b) effective rate