1. Abstract
   1. main technical contribution: Space-Time Appearance Factorization (STAF)
2. intro
   1. Database of Time-Varying Surface Appearance
      1. TSV-BDRF
         1. Time Spatially Varying Bidirectional Reflectance Distribution Function
      2. 26 samples
         1. 1280 (light) measurements of the raw data
         2. the fit BRDF parameters
         3. the estimated STAF results
   2. Space-Time Appearance Factorization (STAF)
      1. advantages
         1. general
         2. data-driven
      2. separates
         1. temporal
            1. estimating a “temporal characteristic curve” in appearance
         2. spatial
   3. Rendering Time-Varying Effects
3. previous work
   1. weathering and corrosion effects
      1. Dorsey et al. 1996;
      2. Dorsey and Hanrahan 1996;
      3. Dorsey et al. 1999;
      4. Merillou et al. 2001;
      5. Chen et al. 2005
   2. Patina(绿绣) formation
      1. based on surface accessibility [Miller 1994]
   3. dust accumulation
      1. based on surface geometry [Hsu and Wong 1995]
   4. wet surfaces
      1. wet surfaces
         1. Jensen et al. [1999]
            1. by combining a reflection model for surface water with subsurface scattering.
   5. measuring appearance changes
      1. Koudelka [2004]
         1. imaged with fixed lighting and a single view
   6. drying on stone
      1. [Lu et al. 2005]
         1. propose a sigmoid model with two spatial parameters
      2. [Jankowsky and Santos 2004]
4. Time-Varying Appearance
   1. TSV-BRDF is a function of 7 dimensions
      1. 2 for spatial location
      2. 2 for incident angle
      3. 2 for outgoing direction
      4. 1 for time variation
5. Acquisition and Database
   1. Acquisition
   2. Parametric Reflectance
      1. non-parametric
         1. used the algorithm in [Vlasic et al. 2003]
         2. performed barycentric interpolation twice
            1. once over view
            2. then over lighting
      2. texture-mapped
         1. simple combination
            1. diffuse Lambertian
            2. simplified Torrance-Sparrow reflectance
   3. Summary and Results
      1. notation p(x, y,t)
6. Modeling and Analysis of Time Variation
   1. deforming a single “temporal characteristic curve” $\rho (t)$
   2. \begin{array} p(x,y,t) & = & A(x,y)\rho (t')+D(x,y) \\ t' & = & R(x,y)t - O(x,y). \end{array}
   3. Data Driven: A,D,R,O,and \rho
      1. \rho(t') – Temporal Characteristic Curve
      2. R(x, y) and O(x, y) – Spatial Rate and Offset
      3. A(x, y) and D(x, y) – Static SV-BRDFs
7. Rendering
   1. Extrapolation
   2. Control
   3. Transfer
   4. Time-Varying Texture Synthesis
   5. The 3D renderings were done using the PBRT package [Pharr and Humphreys 2004]