Transcript Document
Polarization-based Inverse Rendering from Single View Daisuke Miyazaki Robby T. Tan Kenji Hara Katsushi Ikeuchi Modeling cultural assets Geometrical Photometrical Environmental Integrated framework for obtaining 3 types of information 2 Related work Geometry Photometry Environment Tominaga et.al. 2000 Zheng et.al. 1991 Nayar et.al. 1996 Sato et.al. 1999 Ramamoorthi et.al. 2001 Nishino et.al. 2001 Hara et.al. 2002 Proposed method 3 Outline 1. Reflection components separation 2. Shape from polarization using diffuse light 3. Light source estimation from intensity peak 2 4. Reflection parameters estimation by l.s.m. Minimize Ks , σ rendered image real image 4 1. Reflection components separation Dichromatic reflection model Incident light Surface normal Specularly reflected light Diffusely reflected light Air Object 6 Reflection components separation [Tan2002] Input Diffuse •Shape Specular •Illumination •Reflection parameters 7 2. Shape from polarization Related work Koshikawa 1979 Wolff 1990 Rahmann et.al. 2001 Miyazaki et.al. 2002 Proposed method Object Opaque Reflection View Specular 1 Opaque Diffuse Opaque Diffuse Transparent Specular 2 2~5 2 Opaque 1 Diffuse 9 Polarization Incident light Specularly reflected light Diffusely reflected light Air Object 10 Surface normal Camera Polarizer q Azimuth angle Object 11 Azimuth angleφ and intensity difference 255 Imax Intensity Imin 0 1 2 Rotation angle 360 of polarizer Azimuth angle -ambiguity 12 Propagation object Determination of azimuth angle Propagate φ from occluding boundary to inner part of object area (Assumption: smooth surface) [Ikeuchi&Horn1981] Cannot apply to “dimples”(=perfect concave) 13 Zenith angleθ and DOPρ 1 DOP ρ Degree Of Polarization I max I min I max I min 2 n 1 n sin 2 q 2 2 2n 2 n 1 n sin 2 q 4 cos q ρ 0 θ n 2 sin 2 q Zenith angle θ 90° 14 Modification 0.5 DOP ρ Degree Of Polarization 0 2 2 n 1 n Isin max q I min Definition of DOP: 2 2 2 2 2n 2 n 1 n sin 2 qI 4 cos q n sin q max I min I max I min Modified DOP: I max I min u u: modification factor •Raises DOP •Assumption •Closed smooth object •“u” is constant u Zenith angle 90° θ 15 Surface normal φ θ Surface normal 16 Height • Relaxation method [Ikeuchi1984] 2 H H p q x y dxdy 2 Minimize: Surface n p normal q 1 H H q Gradient p y x T 1 p2 q2 Iteratively update: H where, ( i 1) Height H 1 p q ( x, y ) H ( x, y ) 4 x y (i ) 17 3. Illumination estimation Illumination sphere Light source is represented in polar coordinate system (θ, φ) θ=0° L2=(θ2, φ2) L1=(θ1, φ1) L3=(θ3, φ3) φ=90° θ=90° φ=180° θ=90° φ=0° Object φ=270° θ=180° 19 Illumination estimation Detect position of intensity peak Determine light source orientation from the peak 1.Project to (θ, φ)-space 2.Thresholding 3.Detect intensity peak 20 4. Reflection parameters estimation Torrance-Sparrow reflection model Surface Incident normal light Bisector α θi θr View I K d cosθi K s Diffuse reflection Object surface Known: θi, θr, α 1 cosθr e α2 2 2 Specular reflection Unknown: •Diffuse reflection scale; Kd •Specular reflection scale; Ks •Surface roughness; σ 22 Reflection parameters estimation Solve the following least-square problem by steepest-descent method 2 Minimize Ks, σ Ks rendered image 1 cosθr e real image α2 2 2 23 Experimental result Input Azimuth angleφ Intensity I DOPρ 25 Result of shape estimation 26 Result of illumination estimation Actual illumination distribution Estimated illumination distribution 27 Rendering result Input Synthesized image Rendered image under different illumination & view 28 Result for another object Input Synthesized image Estimated shape Rendered image under different illumination & view 29 Conclusions • Estimated geometrical, photometrical, environmental information in one integrated framework – Shape from polarization – Surface reflection parameters from iterative computation – Illumination from intensity peak 30 Application to digital archiving project • Multiple View • Modeling a statue in a room – IBR with • surface normal • reflection parameters Photorealistic preservation 31 Fin (c) Daisuke Miyazaki 2003 All rights reserved. http://www.cvl.iis.u-tokyo.ac.jp/ D. Miyazaki, R. T. Tan, K. Hara, K. Ikeuchi, "Polarization-based Inverse Rendering from Single View," in Proceedings of International Symposium on the CREST Digital Archiving Project, pp.51-65, Tokyo, Japan, 2003.05