Bidirectional texture Function Compression - PowerPoint PPT Presentation

1 / 28
About This Presentation
Title:

Bidirectional texture Function Compression

Description:

It is an extremely difficult and artistic task to model realistic meso- and microstructure ... Homomorphic Factorization [McCool et al. 2001] ... – PowerPoint PPT presentation

Number of Views:52
Avg rating:3.0/5.0
Slides: 29
Provided by: zean
Category:

less

Transcript and Presenter's Notes

Title: Bidirectional texture Function Compression


1
Bidirectional texture Function -
Compression
  • ???

2
Outline
  • Introduction
  • Compression
  • Conclusion

3
Introduction
  • Modeling on different scales

4
Introduction
  • It is an extremely difficult and artistic task to
    model realistic meso- and microstructure
  • -gtUse BTF for meso- and microstructure

5
Introduction
  • BTF is a 6-dimensional texture representation
    which extends the common textures by dependence
    on light- and view-direction Dana et al.
    Transactions on Graphics. 1999

6
Introduction
  • Main Steps

7
Introduction
  • BTF Acquisition
  • Take pictures (spatial dimension) under various
    view and light directions (angular dimensions)

8
Introduction
  • BTF data
  • Collection of discrete textures
  • Apparent BRDF Wong et al. 97
  • Incident direction?surface normal?meso-scale
    shadowing and masking

9
Compression
  • Fitting Analytical Functions
  • Fitting BRDF Model
  • Reflectance Field
  • Statistical Data Analysis
  • Per-Texel Matrix Factorization
  • Full BTF-Matrix Factorization
  • Per-View Factorization
  • Per-Cluster Factorization

10
Fitting BRDF Model - Lafortune Lobes
  • Efficient rendering of spatial bi-directional
    reflectance
  • distribution functions. Mcallister et al.
    Eurographics.
  • 2002
  • Lafortune Lobes Approximate the BRDF by the sum
    of lobes

11
Fitting BRDF Model - Lafortune Lobes
  • Mcallister

12
Fitting BRDF Model Scaled Lafortune Lobes
  • Efficient cloth modeling and rendering. Daubert
    et al. Eurographics. 2001
  • Based on the Lafortune model
  • multiplicative term Tx(v) occlusion

13
Fitting BRDF Model
  • Advantages
  • High compression
  • Efficient evaluation
  • Problems
  • only for a very limited range of materials

14
Reflectance Field
  • Reflectance field based real-time, high-quality
    rendering of bidirectional texture functions
    Meseth et al. ICPR. 2004
  • Surface light field
  • Surface reflectance field
  • Approximate BTF by a set of LF/RF

15
Reflectance Field
  • Advantages
  • High quality
  • Problems
  • Memory cost

16
Compression
  • Fitting Analytical Functions
  • Fitting BRDF Model
  • Reflectance Field
  • Statistical Data Analysis
  • Per-Texel Matrix Factorization
  • Full BTF-Matrix Factorization
  • Per-View Factorization
  • Per-Cluster Factorization

17
Per-Texel Matrix Factorization
  • Apply factorization to the per-textel ABRDFs.
  • Homomorphic Factorization McCool et al. 2001
  • Chained Matrix Factorization Suykens et al.
    Eurographics. 2003
  • Apply chain of factorizations (SVD) to
    reparameterized data
  • Each parameterization accounts for certain
    ABRDF-features

18
Per-Texel Matrix Factorization
  • Advantages
  • Suitable for real-time rendering Combination of
    few factors on GPU
  • Problems
  • Memory consumption
  • Spatial coherence not exploited

19
Full BTF-Matrix Factorization
  • Stack images as column vectors into large matrix

20
Full BTF-Matrix Factorization
  • Write the BTF as sum of products of two functions
  • Number of terms

21
Full BTF-Matrix Factorization
  • TensorTextures Vasilescu et al. Siggraph. 2003
  • Organize BTF-data in 3D-Tensor
  • Apply multi-linear Singular Value Decomposition
    (SVD)

22
Full BTF-Matrix Factorization
  • Advantages
  • Clean and simple
  • Problems
  • Implementation SVD of large BTFs
  • Complex materials require many terms

23
Per-View Factorization
  • Efcient and Realistic Visualization of Cloth
    Sattler et al. EGSR. 2003
  • Apply PCA to slices of the BTF with fixed
    view-direction

24
Per-View Factorization
  • Advantages
  • Suitable for GPU implementation
  • Problems
  • Memory consumption
  • Coherence between different views not exploited

25
Per-Cluster Factorization
  • Local-PCA with clustering along spatial dimension
    leads to

26
Per-Cluster Factorization
  • Advantages
  • Suitable for GPU implementation
  • Exploits redundancy across spatial and angular
    dimensions
  • Problems
  • Expensive fitting

27
Conclusion
28
Conclusion
  • Quality Linear basis decomposition is better
    than parametric reflectance models
  • Best quality Per-View Factorization
  • Best space compression Per-Cluster Factorization
Write a Comment
User Comments (0)
About PowerShow.com