Presentation is loading. Please wait.

Presentation is loading. Please wait.

Computer Graphics Laboratory, Hiroshima City University All images are compressed.

Similar presentations


Presentation on theme: "Computer Graphics Laboratory, Hiroshima City University All images are compressed."— Presentation transcript:

1 http://www.hiroshima-cu.ac.jp/ Computer Graphics Laboratory, Hiroshima City University All images are compressed.

2 http://www.hiroshima-cu.ac.jp/ Computer Graphics Laboratory, Hiroshima City University Interactive shadow removal from a single image using hierarchical graph cut Daisuke Miyazaki Yasuyuki Matsushita Katsushi Ikeuchi

3 http://www.hiroshima-cu.ac.jp/ CG Lab, Hiroshima City University Motivation

4 http://www.hiroshima-cu.ac.jp/ CG Lab, Hiroshima City University Prior work multiple single automatic interactive Weiss 01; Matsushita et al. 04 Finlayson et al. 04; Fredembach & Finlayson 06; Tappen et al. 05; Baba et al. 04; Arbel & Hel-Or 07; Nielsen & Madsen 07 Wu et al. 07

5 http://www.hiroshima-cu.ac.jp/ CG Lab, Hiroshima City University Contribution Prior Smoothness Graph cut Alpha expansion User satisfaction Robust Intuitive Fast

6 http://www.hiroshima-cu.ac.jp/ CG Lab, Hiroshima City University Algorithm flow InputOver-segmentationLazy snapping Shadow removal InitialInteractionOutput 1 1 2 2 3 3 4 4 5 5 6 6

7 http://www.hiroshima-cu.ac.jp/ CG Lab, Hiroshima City University Shadow matte Prior of shadowless image Prior of shadow image Smoothness of shadowless image Smoothness of shadow image Solve

8 http://www.hiroshima-cu.ac.jp/ CG Lab, Hiroshima City University Prior of shadowless image Likelihood P appropriatelargesmall

9 http://www.hiroshima-cu.ac.jp/ CG Lab, Hiroshima City University Prior of shadow image smalllargeappropriate L0.7-NormL1-NormL2-Norm

10 http://www.hiroshima-cu.ac.jp/ CG Lab, Hiroshima City University Smoothness of shadowless image smalllargeappropriate

11 http://www.hiroshima-cu.ac.jp/ CG Lab, Hiroshima City University Smoothness of shadow image smalllargeappropriate

12 http://www.hiroshima-cu.ac.jp/ CG Lab, Hiroshima City University Smoothness shadow 0 1 Interactive parameter optimization Smoothness shadowless Prior shadowless Prior shadow 0 0.25 0 0.4 0.7 Init Initial 1st 2nd 3rd 4th 5th 6th

13 http://www.hiroshima-cu.ac.jp/ CG Lab, Hiroshima City University Approach based on alpha expansion Alpha expansionHierarchical graph cut alpha=2 alpha=2,4,... alpha=4 3 15 2 4 1 1 3 5 5 1 1 3 5 5 3 51 2 1 2 3 5 4 5 4 1 2 1 2 5 4

14 http://www.hiroshima-cu.ac.jp/ CG Lab, Hiroshima City University Hierarchical structure 32 1648 8244056 8 162432404856 163248 8 162432404856 163248 20

15 http://www.hiroshima-cu.ac.jp/ CG Lab, Hiroshima City University Graph construction example paq  

16 http://www.hiroshima-cu.ac.jp/ CG Lab, Hiroshima City University Algorithm Alpha expansionHierarchical graph cut 1.A = {{0}, {32}, {16}, {48}, {8, 40}, {24, 56}, {4, 20, 36, 52}, {12, 28, 44, 60},..., {3, 7, 11,..., 63}} 2.for i = 0 to 11 do a.for all nodes&edges do I. II.add nodes&edges using & b.solve max-flow/min-cut problem 3.Iterate 2 until convergence 1.A = {0, 1, 2,..., 63} 2.for i = 0 to 63 do a.for all nodes&edges do I. II.add nodes&edges using & b.solve max-flow/min-cut problem 3.Iterate 2 until convergence

17 http://www.hiroshima-cu.ac.jp/ CG Lab, Hiroshima City University Output of Hierarchical graph cut Stereo matching [Art] Stereo matching [Laundry] Image restoration Shadow removal InputGround truthIshikawa 2003  expansion Hierarchical cut

18 http://www.hiroshima-cu.ac.jp/ CG Lab, Hiroshima City University Speed of hierarchical graph cut Stereo matching [Art] Stereo matching [Laundry] Image restoration Shadow removal Speed-up  6.8  11.4  16.6  3.4 Error difference+4.6%+3.0%-0.4%+0.0% Hierarchical graph cut Alpha expansion [sec] Iteration 250 7 Energy 0 1.1  10 7 1.7  10 6 1.5  10 6

19 http://www.hiroshima-cu.ac.jp/ CG Lab, Hiroshima City University Incremental improvement InputTruthFinlayson 2004Wu 2007InitialInteractive [Ours] Error Number of strokes for parameter optimization50 Finlayson 2004 Wu 2007 Ours

20 http://www.hiroshima-cu.ac.jp/ CG Lab, Hiroshima City University 0 strokes (param. opt.) 1.4 sec  1 strokes2.1 sec  6 strokes Natural images 1.2 sec  4 strokes

21 http://www.hiroshima-cu.ac.jp/ CG Lab, Hiroshima City University 2.0 sec  12 strokes1.5 sec  2 strokes Natural images 1.9 sec  15 strokes1.5 sec  8 strokes

22 http://www.hiroshima-cu.ac.jp/ CG Lab, Hiroshima City University 5.8 sec  27 strokes2.8 sec  40 strokes Natural images 6.3 sec  32 strokes2.0 sec  26 strokes

23 http://www.hiroshima-cu.ac.jp/ CG Lab, Hiroshima City University Aerial images

24 http://www.hiroshima-cu.ac.jp/ CG Lab, Hiroshima City University Conclusions Hierarchical graph cut Cost functionUser interaction

25 http://www.hiroshima-cu.ac.jp/ CG Lab, Hiroshima City University Discussions Cost function User interaction Hierarchical graph cut

26 http://www.hiroshima-cu.ac.jp/ Computer Graphics Laboratory, Hiroshima City University (c) Daisuke Miyazaki 2009 All rights reserved. http://www.cg.info.hiroshima-cu.ac.jp/ Daisuke Miyazaki, Yasuyuki Matsushita, Katsushi Ikeuchi, "Interactive shadow removal from a single image using hierarchical graph cut," Lecture Notes in Computer Science (Proc. Asian Conference on Computer Vision), vol. 5994, part 1, 2009.


Download ppt "Computer Graphics Laboratory, Hiroshima City University All images are compressed."

Similar presentations


Ads by Google