1 / 22

Evolutionary Computation and Image Re-Coloring

Evolutionary Computation and Image Re-Coloring. Gary R. Greenfield Mathematics & Computer Science University of Richmond Computational Aesthetics in Graphics, Visualization, and Imaging Dagstuhl Seminar, 2006. Motivation.

gratia
Télécharger la présentation

Evolutionary Computation and Image Re-Coloring

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Evolutionary Computation and Image Re-Coloring Gary R. Greenfield Mathematics & Computer Science University of Richmond Computational Aesthetics in Graphics, Visualization, and Imaging Dagstuhl Seminar, 2006

  2. Motivation • Is there a role for evolutionary computation (EC) in computational aesthetics? • Example: Image Re-coloring.

  3. Background (1 of 4) • E. Reinhard, M. Ashikhmin, B. Gooch, P. Shirley, Color transfer between images, IEEE CG&A, 2001, 34-41. • B. Meier, A. Spalter, D. Karelitz, Interactive color palette tools, IEEE CG&A, 2004, 64-72. • G. Greenfield, D. House, Palette-driven color transfer, Computational Aesthetics 2005, 91-99. • M. Grundland, N. Dodgson, Color search and replace, Computational Aesthetics 2005, 100-109.

  4. Background (2 of 4) • L. Neumann, A. Neumann, Color style transfer techniques using hue, lightness and saturation histogram matching, Computational Aesthetics 2005, 111-122. • A. Hertzmann, C. Jacobs, N. Oliver, B. Curless, D. Salesin, Image Analogies, Proc. SIGGRAPH ’01, 2001, 327-340. • T. Welsh, M. Ashikhmin, K. Mueller, Transferring color to grayscale images, Proc. SIGGRAPH ’02, 2002, 277-280. • Y. Chang, S. Saito, M. Nakajima, A framework for transfer colors based on the basic color categories, Proc. Computer Graphics International, 2003, 176-181.

  5. Background (3 of 4) • L. Neumann, A. Nemcsics, A. Neumann, Computational color harmony based on Coloroid system, Computational Aesthetics 2005, 231-240. • A. Levin, D. Lischinski, Y. Weiss, Colorization using optimization, Proc. SIGGRAPH ’04, 2004, 689-694. • R. Irony, D. Cohen-Or, D. Lischinski, Colorization by example, Rendering Techniques, 2005, 201-210. • A. Gooch, S. Olsen, J. Tumblin, B. Gooch, Color2Gray: salience preserving color removal, Proc. SIGGRAPH ’05, 2005, 634-639.

  6. Background (4 of 4) • G. Greenfield, An algorithmic palette tool, UR Technical Report, 1994. • J. Barallo, V. de Spinadel, Colouring algorithms and fractal art, J. of Math. and Design, 2001, 27-32. • C. Farsi, C. Collins, Visual synthesis and esthetic values, Proc. Generative Arts Conference, 2004, 40-45. • G. Greenfield, Image re-coloring using multi-objective optimization, Proc. Art+Math=X Conference, 2005, 83-87.

  7. Color Look-Up Tables • Fix a “small” set S = {C1,…,CM} consisting of M <HSV | RGB | YIQ |…> colors. • A color look-up table (CLUT) of length L is a list T = (c1,…,cL) of not necessarily distinct colorschosen from S. • The image I can be re-colored by T if each pixel P can be identified with some integer I such that 1 <= I <= L (viz. re-color P using color cI).

  8. Evolving CLUT’s • One way to evolve CLUTs is by randomly initializing CLUT segments and then using crossover as the recombination operator coupled with mutation operators such as: • “Flow” • “Extend” • “Copy” • “Blot”

  9. An Evaluation Framework • Unfortunately, the problem of how to evaluate the aesthetic quality of a CLUT re-coloring still remains. • The following re-colorings were done by applying evolutionary multi-objective optimization (EMO) using certain color and geometric characteristics of low res segmented versions of the re-colored images. For example, F1(I) = A2,6 Jm/2,m + C4,5 F2(I) =min(T1,4.2,T2,3.7) B1,4

  10. Re-Coloring Results

  11. The Rogue’s Gallery

  12. Lessons Learned • The fewer colors that are required, the more successful the re-coloring will be. • Images with many large, juxtaposed, and differently colored regions “confuse” the re-coloring algorithm. • Luminance criteria should be taken into consideration.

  13. Other Ideas and Suggestions • Introduce penalty terms for prohibiting color neighbor mismatches in the re-coloring. • Define CLUT’s as discretizations of continuous curves (e.g. evolve the spline points for such curves). • Adopt a SETI Approach – Use on-line voting schemes to evolve re-colorings… and learn principles and preferences?!

More Related