1,a) 1,b) 1,c) 1,d) 2,e) 2,f) 2,g) 1. [1] [2] 2 [3] 1 599 8531 1 1 Osaka Prefecture University 1 1, Gakuencho, Naka, Sakai, Osaka 599 8531, Japan 2 565 0871 Osaka University 1 1, Yamadaoka, Suita, Osaka 565 0871, Japan a) nakamura@m.cs.osakafu-u.ac.jp b) yuzuko@cs.osakafu-u.ac.jp c) masa@cs.osakafu-u.ac.jp d) kise@cs.osakafu-u.ac.jp e) makihara@am.sanken.osaka-u.ac.jp f) muramatsu@am.sanken.osaka-u.ac.jp g) yagi@am.sanken.osaka-u.ac.jp [3][4] [5] [6] [5] c 2016 Information Processing Society of Japan 1
2. Level Sets[7] [8] Level Sets Level Sets Rother Grab Cut[9] Grub Cut Li Lazy Snapping[10] Lazy Snapping watershed [11] [5] Grab Cut Lazy Snapping [5] [6] [5] [5] 3.1 3.2 3.1 1 DP 1 3 1 3. [5] 1 [5] [6] [5] 2 2 c 2016 Information Processing Society of Japan 2
3 5 DP 4 [12] 4 4 M m (m = 1,..., M) g m (ϕ, S) R W H W H S = [s, s x, s y ] T ϕ = (ϕ 1, ϕ 2,..., ϕ l ) s = (s 1, s 2,..., s r ) s x = (s x1, s x2,..., s xt ) x s y = (s y1, s y2,..., s yv ) y L L = lrtv n (n = 1,..., N) f(n) R W H x, y I(g m (ϕ, S), x, y)i(f(n), x, y) f(n) g m (ϕ, S) d 1 (x,y) d 1 = 1 min{ I(f(n), x, y), I(g m (ϕ, S), x, y) } (x,y) max{ I(f(n), x, y), I(g m (ϕ, S), x, y) } (1) DP DP 5 DP DP 3.2 q (X q = BG)q (X q = F G) E(X) = g q (X q ) + h pq (X p, X q ) + g s (X q ) q Q (p,q) E q Q (2) X Q X q E 1 2 3 [5] (2) [8] c 2016 Information Processing Society of Japan 3
[5] 3.2.1 q (X q = BG) q q (X q = F G) q q µ bg,q Σ bg,q q c q d bg,q d bg,q = (c q µ bg,q ) T Σ 1 bg,q (c q µ bg,q ) (3) q g q (X q = BG) = ω bg exp( κ bg d bg,q ) (4) ω bg κ bg q µ bg,q k k k µ fg,k Σ fg,k q c q d fg,k d fg,k = (c q µ fg,k ) T Σ 1 fg,k (c q µ fg,k ) (5) g q (X q = F G) g q (X q = F G) = ω fg exp( κ fg min k d fg,k ) (6) ω fg κ fg 3.2.2 pq c p c q 0 (X p = X q ) h pq (X p, X q ) = c ω sm exp( κ q c p 2 sm c q+c p 2 +ε ) (X p X q ) (7) ω sm,κ sm,ε 3.2.3 3.1 d Q,sh d Q,sh 1 g sh (X q = BG) = ω sh 1 + exp( κ sh d q,sh ) 1 g sh (X q = F G) = ω sh 1 + exp(κ sh d q,sh ) (8) (9) κ sh,ω sh 4. [5] 6(a) 6(b) [12] [5] 7 (1) c 2016 Information Processing Society of Japan 4
(a) A (b) B 8 (a) (b) 6 1 [] 800 450 [fps] 12 [] 100( A) 80( B) 7 x, y I(g m (ϕ, S), x, y)i(f(n), x, y) f(n) g m (ϕ, S) d 2 ω(x, y) (x,y) d 2 = 1 ω(x, y)min{ I(f(n), x, y), I(g m (ϕ, S), x, y) } (x,y) ω(x, y)max{ I(f(n), x, y), I(g m (ϕ, S), x, y) } (10) (10) 5. [5] [5] [5] 6 1 2141 1 913 A 5 9 DP (10) 20% ω(x, y) = 7.5 ω(x, y) = 1.0 i u e e = i (11) u Gait energy image[4] [1] A B A B A Probe B Gallery 50 50 50 5.1 2 8 1 A B 56 50 9 c 2016 Information Processing Society of Japan 5
2 [%] 1 0.291 0.169 0.147 2 0.091 0.089 0.088 3 0.161 0.156 0.155 4 0.206 0.187 0.190 5 0.172 0.173 0.169 0.184 0.154 0.149 11 3 3 [%] 62 70 64 (a) (b) 10 5.2 2 2 5.2 5.3 30% 11 4 6. c 2016 Information Processing Society of Japan 6
4 [%] 60 62 60 JSPS (A)JP15H01693 No. 3, ACM, pp. 309 314 (2004). [10] Li, Y., Sun, J., Tang, C.-K. and Shum, H.-Y.: Lazy snapping, ACM Transactions on Graphics (ToG), Vol. 23, No. 3, pp. 303 308 (2004). [11] Vincent, L. and Soille, P.: Watersheds in digital spaces: an efficient algorithm based on immersion simulations, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 13, No. 6, pp. 583 598 (1991). [12] Sloan, K. R. and Tanimoto, S. L.: Progressive refinement of raster images, IEEE Transactions on Computers, Vol. 28, No. 11, pp. 871 874 (1979). [1] (CVIM) Vol. 2013-CVIM-186, No. 3, pp. 1 10 (2013). [2] Vol. 43, pp. 21 25 (2016). [3] (CVIM) Vol. 2013-CVIM-187, No. 10, pp. 1 8 (2013). [4] Han, J. and Bhanu, B.: Individual recognition using gait energy image, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 28, No. 2, pp. 316 322 (2006). [5] Makihara, Y., Tanoue, T., Muramatsu, D., Yagi, Y., Mori, S., Utsumi, Y., Iwamura, M. and Kise, K.: Individuality-preserving silhouette extraction for gait recognition, IPSJ Transactions on Computer Vision and Applications, Vol. 7, pp. 74 78 (2015). [6] Makihara, Y. and Yagi, Y.: Silhouette extraction based on iterative spatio-temporal local color transformation and graph-cut segmentation, Proceedings of 19th International Conference on Pattern Recognition, pp. 1 4 (2008). [7] Sussman, M., Smereka, P. and Osher, S.: A level set approach for computing solutions to incompressible twophase flow, Journal of Computational Physics, Vol. 114, No. 1, pp. 146 159 (1994). [8] Boykov, Y. Y. and Jolly, M.-P.: Interactive graph cuts for optimal boundary & region segmentation of objects in ND images, Proceedings. Eighth IEEE International Conference on Computer Vision, Vol. 1, pp. 105 112 (2001). [9] Rother, C., Kolmogorov, V. and Blake, A.: Grabcut: Interactive foreground extraction using iterated graph cuts, ACM Transactions on Graphics (TOG), Vol. 23, c 2016 Information Processing Society of Japan 7