Fig. 1. Pixel filling principle

Fig. 2. Determine missing edges by EPGVF field. (a) Test image 2; (b) edge gradient field image; (c) EPGVF field image; (d) edge energy image; (e) edge fracture mark image

Fig. 3. Edge recovery by Snake algorithm. (a) Initial interpolation image of edge fitting; (b) image after iterative fitting; (c) restored image

Fig. 4. Experimental results of proposed method. (a) PCBA edge map; (b) differentiation of x direction of edge gradient field; (c) differentiation of y direction of edge gradient field; (d) differentiation of x direction of EPGVF field; (e) differentiation of y direction of EPGVF field; (f) EPGVF gradient field; (g) edge step energy map; (h) edge mark image of proposed method

Fig. 5. Test images of anisotropic dense edge separation. (a)(g) Local enlarged images of test images; (b)(h) test images; (c)(i) results of traditional algorithm; (d)(j) results of algorithm in Ref. [8]; (e)(k) results of algorithm in Ref. [9]; (f)(l) results of proposed method

Fig. 6. Processing results of several algorithms. (a) Original image; (b) traditional algorithm; (c) algorithm in Ref. [8]; (d) algorithm in Ref. [9]; (e) proposed method

Fig. 7. Unlabeled image and labeled image of connected domain. (a) Unlabeled image; (b) labeled image