Color Constancy from Illumination Changes

Transcription

1 (MIRU2004) E {rei,robby,ki}@cvl.iis.u-tokyo.ac.jp Finlayson [10] Color Constancy from Illumination Changes Rei KAWAKAMI,RobbyT.TAN, and Katsushi IKEUCHI Institute of Industrial Science, University of Tokyo Komaba, Meguro-ku, Tokyo, , JAPAN {rei,robby,ki}@cvl.iis.u-tokyo.ac.jp Abstract Elimination of illumination color in color images is essential for various fields in computer vision. Many methods have been proposed to solve it, however there are few methods employing varying illumination as a constraint. We found Finlayson et al. s method [10] applicable for natural images, and extended it by making the reference illumination variable. To evaluate our method, we have conducted a number of experiments with natural images and also with textured images mapped on 3D data. The evaluation shows that the method has higher accuracy and robustness compared to the previous method. Key words color, color constancy, varying illumination, Planckian locus, texture 1. SPD(Spectral Power Distribution) Color constancy Color constancy [3], [5], [6], [9], [11] [16], [19] [24] Dichromatic based Diffuse based Dichromatic based [5], [9], [12] [14], [19], [20], [22] Diffuse based [3], [6], [11], [21], [23], [24] Diffuse-based [11], [21], [23], [24] [2], [4], [10]

2 D zmura [4] Finlayson [10] Barnard [2] Retinex Algorithm [18] Finlayson Finlayson [10] Finlayson I c (1) I c = S(λ)E(λ)q c(λ)dλ (1) Ω S(λ) E(λ) q c c (R,G,B) (Ω) Dirac (1) (2) 2. 2 (Planckian locus) ( ) Planckian locus Planckian locus [12], [17] Planckian locus RGB SPD(Spectral Power Distribution) Planck ( (6)) M(λ) =c 1λ 5 [exp(c 2/λT) 1] 1 (6) c 1,c (Wm 2 ), (mk), λ (m) T (K). (7) SPD RGB I c = M(λ, T )q c(λ)dλ (7) Ω Planckian locus I c = S ce c (2) [1], [7], [8] RGB RGB (3) 1 Planckian locus σ c = Ic I B (3) c R,G (2) σ c (4) σ c = s ce c (4) s c,e c S c,e c (4) (5) σc,σ 1 c 2 e 1,e 2 [ ] [ ][ σr 2 e 2 r/e 1 r 0 = σg 2 0 e 2 g/e 1 g σ 1 r σ 1 g ] (5) 2. 3 Finlayson [10] Finlayson [10] Finlayson 3. e ref e ref Φ Φ (8) [ ] e ref r /r 0 Φ (8) 0 e ref g /g ( (5)) Planckian locus (1/r, 1/g) 2

3 2 Planckian locus 3. Finlayson 5 σ 1 σ 2 Φσ 1 Φσ 2 3 Judd Daylight Phases CIE Finlayson Judd Daylight Phases(D48,D55,D65,D75,D100) CIE (A,B,C) ( 3 [17]) Planckian locus Planckian locus 1/r 1/g Φ σ Φσ σ 1,σ 2 Φσ 1,Φσ 2 (r, g) =(e 1 r,e 1 g) (e 2 r,e 2 g) σ ref σ ref (9) (0,0) 5 Planckian locus ( (8) e ref ) 3. 1 Finlayson s σ 1 σ 2 ( 6) Φσ 1 Φσ 2 = σ ref (9) s s Planckian locus Planckian locus ( 1) Planckian locus e r,e g s s r,s g (4) σ r,σ g σ 1,σ 2 ( 7 Ex.3,Ex.4) s

4 σ real σ real σ real 8 Finlayson s 8 s s 3. g 1 g g 3. 3 (σ 1 σ 2 ) (σ ref ) σ 1 σ ref σ 2 ( (8) e ref ) se ref Ex.1 Ex.2 7 Ex.3 Ex RGB e ref 90 e ref (90 ) g g δ err Planckian locus 100(k) 0.01 CIE 0.2 [25] CCD SONY DXC Photo Research PR-650 (Photo Research Reflectance Standard model SRS-3) Finlayson 15:00 18:00( )

5 9 Finlayson ( ) r g b r g b a 9.b 9.c 9.d 9.e Finlayson Finlayson Finlayson [25] :00( ) 12 13, 14 RGB ( + ) CIE (σ c = I c/σi i) Finlayson Finlayson 0.32 R B G Finlayson G 1 (CIE )

6 Finlayson [1] K. Barnard, F. Ciurea, and B. Funt. Sensor sharpening for computational color constancy. Journal of Optics Society of America A., 18(11): , [2] K. Barnard, G. Finlayson, and B. Funt. Color constancy for scenes with varying illumination. Computer Vision and Image Understanding, 65(2): , [3] D.H. Brainard and W.T. Freeman. Bayesian color constancy. Journal of Optics Society of America A., 14(7): , [4] M. D Zmura. Color constancy: surface color from changing illumination. Journal of Optics Society of America A., 9(3): , [5] M. D Zmura and P. Lennie. Mechanism of color constancy. Journal of Optics Society of America A., 3(10): , [6] G.D. Finlayson. Color in perspective. IEEE Trans. on Pattern Analysis and Machine Intelligence, 18(10): , [7] G.D. Finlayson. Spectral sharpening: what is it and why is it important. In The First European Conference on Colour in Graphics, Image and Vision, pages , [8] G.D. Finlayson, M.S. Drew, and B.V. Funt. Spectral sharpening sensor transformations for improved color constancy. Journal of Optics Society of America A., 11(10): , [9] G.D. Finlayson and B.V. Funt. Color constancy using shadows. Perception, 23:89 90, [10] G.D. Finlayson, B.V. Funt, and K. Barnard. Color constancy under varying illumination. in proceeding of IEEE International Conference on Computer Vision, pages , [11] G.D. Finlayson, S.D. Hordley, and P.M. Hubel. Color by correlation: a simple, unifying, framework for color constancy. IEEE Trans. on Pattern Analysis and Machine Intelligence, 23(11): , [12] G.D. Finlayson and G. Schaefer. Solving for color constancy using a constrained dichromatic reflection model. International Journal of Computer Vision, 42(3): , [13] G.D. Finlayson and S.D.Hordley. Color constancy at a pixel. Journal of Optics Society of America A., 18(2): , [14] B.V. Funt, M. Drew, and J. Ho. Color constancy from mutual reflection. International Journal of Computer Vision, 6(1):5 24, [15] J.M. Geusebroek, R. Boomgaard, S. Smeulders, and H. Geert. Color invariance. IEEE Trans. on Pattern Analysis and Machine Intelligence, 23(12): , [16] J.M. Geusebroek, R. Boomgaard, S. Smeulders, and T. Gevers. A physical basis for color constancy. In The First European Conference on Colour in Graphics, Image and Vision, pages 3 6, [17] D.B. Judd, D.L. MacAdam, and G. Wyszecky. Spectral distribution of typical daylight as a function of correlated color temperature. Journal of Optics Society of America, 54(8): , [18] E.H. Land and J.J. McCann. Lightness and retinex theory. Journal of Optics Society of America, 61(1):1 11, [19] H.C. Lee. Method for computing the scene-illuminant from specular highlights. Journal of Optics Society of America A., 3(10): , [20] H.C. Lee. Illuminant color from shading. In Perceiving, Measuring and Using Color, page 1250, [21] C. Rosenberg, M. Hebert, and S. Thrun. Color constancy using kl-divergence. In in proceeding of IEEE Internation Conference on Computer Vision, volume I, page 239, [22] R. T. Tan, K. Nishino, and K. Ikeuchi. Illumination chromaticity estimation using inverse intensity-chromaticity space. in proceeding of IEEE Conference on Computer Vision and Pattern Recognition, [23] S. Tominaga, S. Ebisui, and B.A. Wandell. Scene illuminant classification: brighter is better. Journal of Optics Society of America A., 18(1):55 64, [24] S. Tominaga and B.A. Wandell. Natural scene-illuminant estimation using the sensor correlation. Proceedings of the IEEE, 90(1):42 56, [25]. Simultaneous registration of 2d images onto 3d models for texture mapping., 2003.