Title: 3D Geometry Coding using Mixture Models and the Estimation Quantization Algorithm
13D Geometry Coding usingMixture Models andthe
Estimation Quantization Algorithm
- Sridhar Lavu
- Masters Defense
- Electrical Computer Engineering
DSP Group
Rice University
September 2002
23D Surfaces
- Video games
- Animations - Bugs Life, Toy Story 2
- 3D object modeling - CAD
- e-commerce
33D Surfaces
- Geometry, color, texture
- 3D scanning
- Polygon meshes
- Problem - large data sets
- Geometry compression
100,000 triangles
4Contribution
- 3D geometry coder
- Multilevel representation
- Normal meshes
- EQ algorithm
- Estimation-Quantization (EQ)
- Local context information
- RD optimization
5Related Work
- Zerotree coder for the wavelet coefficients of
normal meshes - RD optimization based quantization algorithm for
the wavelet coefficients of meshes
6Outline
- 3D surface data
- Multilevel representation
- Normal meshes
- Wavelet transform
- EQ algorithm
- Error metrics
- Results
73D geometry data
- Geometry
- Polygon meshes
- Geometry connectivity
Geometry 0.0 0.0 0.0 1.0 0.0 0.0 1.0 1.0 0.0 0.0
1.0 0.0 0.5 0.5 1.0
Connectivity 0 1 2 2 3 1 0 1 4 1 2 4 2 3 4 3 0 4
8Multilevel Representations
Original
Coarse
Multilevel triangular meshes
Original ? Normal meshes
9Normal meshes
- Multilevel representation
- Base mesh
- Successively refine the mesh
- Subdivision
10Subdivision
- Linear subdivision
- Butterfly subdivision
- Loop subdivision
11Butterfly Subdivision
12Normal Meshes
- Predict b and n
- Find intersection
- Store offset
- 1 number per vertex
13Wavelet Transforms
- Irregular data
- Lifting scheme predict and update
- Subdivision predict step
- Wavelet transforms
- Butterfly wavelet transform
- Loop wavelet transform
14Wavelet Transforms and Normal Meshes
Wavelet coefficients
Non-normal vertices
15Related Work - Zerotree
Zerotrees
Mesh zerotree coding
Mesh zerotree
Zerotree coding
EQ coding
16Review
- Multilevel representations for meshes
- Normal meshes
- Wavelet transforms
- Subdivision
- Lifting
- Related work - ZT based algorithm
- Contribution EQ based algorithm
173D EQ Coder
- Local context information
- Model for wavelet coefficients
- Generalized Gaussian distribution
- EQ Algorithm
- Estimate Step
- Quantize Step
- RD Optimization
18Wavelet Coefficient Model
- Generalized Gaussian distribution
19Wavelet Coefficient Model
- Generalized Gaussian (GGD)
- ? Shape Fixed at each level
- ? Variance Local neighborhood
- ? Mean Zero
20EQ Algorithm
- Scan the vertices
- Estimate, quantize, encode
- Estimate step - variance
- Local neighborhood
- Causal neighborhood
- Quantized neighbors
- Quantize step
- Deadzone quantizer
- RD optimization
21EQ Algorithm (cont.)
- RD optimization
- Rate -log(probability)
- Distortion MSE of coefficients
- Entropy coding
- Arithmetic coder
22Normal vs. Tangential
- Smooth surfaces
- Global error contribution
- Normal Higher
- Tangential Lower
- Precision
- Normal Higher Lower l
- Tangential Lower Higher l
- Most tangential components are zero
- Single quantizer per level
23Neighborhood
24Ordering - Base Triangles
25Ordering - Vertices
26Summary of EQ Algorithm
- Pick l
- Determine ordering
- Ordering of base triangles
- Ordering inside each base triangle
- Local causal neighborhood
- Estimate s
- Quantize using RD optimization
- Normal vs. tangential
27Performance Measure
- Error metrics
- MSE ?
- Hausdorff distance
- Min, max, mean, mean squared
28Results
- Metric - PSNR
- Bits-per-vertex (bpv)
- Reconstructed mesh vs. original mesh
- Metro and MeshDev software tools
29Results - EQ vs. ZT
30Results EQ vs. ZT(Lifted Butterfly)
31Results - EQ vs. ZT(Loop Wavelets)
32Results (Bounds)
- Upper bound
- Complete context
- Lower bound
- No context
33Summary
- Multilevel representations
- Normal meshes
- Wavelet transforms
- GGD model
- Local context based coder
- EQ vs. ZT
34Conclusion Future Work
- Conclusions
- GGD model EQ algorithm
- 0.5 1 dB gain
- Future work
- Vertex based error for RD optimization
- New algorithms
- Space-Frequency quantization (SFQ)
35(No Transcript)
36Scaling Coefficients andConnectivity
- Scaling coefficients
- Vertices of base mesh
- Uniform quantization
- Connectivity
- Semi-regular connectivity
- Base mesh connectivity
- TG Coder (lossless)
37Lifting (Predict, Update)
Forward
Inverse
38Lifting - Haar
39Loop Wavelet Transform
40Causal Neighborhoods
41EQ Unpredictable sets
- Empty causal neighborhood
- Zero s estimate
- Classify as unpredictable (U) set
- Model U set as zero-mean GGD
- Use a single s and n for U set
42EQ Threshold step
- Iteration of E and Q steps
- First iteration
- Threshold coefficients
- Partition U and P sets
- Estimate s and n
- Use estimates in next iteration
43Normal Predictable Set
44Normal Unpredictable Set
45Tangential Set
46Hausdorff Distance
47Mesh Zerotree Coding
48Results Venus PSNR
BPV 0.25 0.5 1.0 2.0 3.0 4.0
EQ lifted BW 63.7 68.6 74.2 79.2 81.7 83.2
ZT lifted BW 63.0 68.2 73.7 78.9 81.7 81.9
EQ unlifted BW 63.5 68.6 74.1 78.9 81.4 83.0
ZT unlifted BW 62.4 67.8 73.0 78.4 81.2 81.5
EQ Loop Wavelet 60.0 65.3 71.3 76.4 79.4 81.4
ZT Loop Wavelet 60.9 66.1 71.8 77.1 79.7 79.7
49Results Rabbit PSNR
BPV 0.25 0.5 1.0 2.0 2.5 3.0
EQ lifted BW 70.3 75.7 80.9 84.2 85.1 85.6
ZT lifted BW 69.3 75.1 80.9 84.0 84.1 84.1
EQ unlifted BW 70.0 75.3 80.6 84.0 85.0 85.5
EQ unlifted BW 68.7 74.7 80.4 83.6 83.6 83.6