GenSSS: A Genetic Algorithm for Measured Subsurface Scattering Representation
(Image Comparison)

Below we compare our subsurface scattering model against the reference image, Peers et al.'s [2006] subsurface scattering model, and SubEdit model [2009] at comparable data sizes. We also visually show the effect of adjusting the K term (i.e., the total number of terms) in our GenSSS representation. For higher disparity, the color-coded difference images were scaled by 10. Furthermore, we show response image and albedo image comparisons of well-known subsurface scattering representations at comparable data sizes. Finally, we show results from our application, which can be used to convert heterogeneous materials into homogeneous material representations.

Tip: Move the mouse over the images to compare the algorithms on specific parts of the scene.
Warning: This page include many full-resolution images, so the page may take some time to fully load.

◄ back to the publication web page

Jump to comparisons:

K term rep. K term diff. SSS rep. SSS diff. Response img. Albedo img. Application

Comparison of our GenSSS representation with various K terms

The visual comparisons of our subsurface scattering representation with various K parameters. The scenes were rendered at 16 samples / pixel. From top to bottom: artificial stone, chessboard (8×8), marble (close up) and yellow wax materials on lucy, buddha, kitten and dragon objects, respectively:

Reference Our GenSSS model, <em>K</em>=1 Our GenSSS model, <em>K</em>=2 Our GenSSS model, <em>K</em>=5
Reference
Our GenSSS model, K=1
Our GenSSS model, K=2
Our GenSSS model, K=5
Reference Our GenSSS model, <em>K</em>=1 Our GenSSS model, <em>K</em>=2 Our GenSSS model, <em>K</em>=5
Reference
Our GenSSS model, K=1
Our GenSSS model, K=2
Our GenSSS model, K=5
Reference Our GenSSS model, <em>K</em>=1 Our GenSSS model, <em>K</em>=2 Our GenSSS model, <em>K</em>=5
Reference
Our GenSSS model, K=1
Our GenSSS model, K=2
Our GenSSS model, K=5
Reference Our GenSSS model, <em>K</em>=1 Our GenSSS model, <em>K</em>=2 Our GenSSS model, <em>K</em>=5
Reference
Our GenSSS model, K=1
Our GenSSS model, K=2
Our GenSSS model, K=5

Comparison of difference images with various K terms

The visual comparisons of the color-coded difference images of our GenSSS representation with various K parameters. From top to bottom: artificial stone, chessboard (8×8), marble (close up) and yellow wax materials on lucy, buddha, kitten and dragon objects, respectively:

Our GenSSS model, <em>K</em>=1 Our GenSSS model, <em>K</em>=2 Our GenSSS model, <em>K</em>=5
Our GenSSS model, K=1
Our GenSSS model, K=2
Our GenSSS model, K=5
Our GenSSS model, <em>K</em>=1 Our GenSSS model, <em>K</em>=2 Our GenSSS model, <em>K</em>=5
Our GenSSS model, K=1
Our GenSSS model, K=2
Our GenSSS model, K=5
Our GenSSS model, <em>K</em>=1 Our GenSSS model, <em>K</em>=2 Our GenSSS model, <em>K</em>=5
Our GenSSS model, K=1
Our GenSSS model, K=2
Our GenSSS model, K=5
Our GenSSS model, <em>K</em>=1 Our GenSSS model, <em>K</em>=2 Our GenSSS model, <em>K</em>=5
Our GenSSS model, K=1
Our GenSSS model, K=2
Our GenSSS model, K=5

Comparison of subsurface scattering representations

The visual comparisons between the SubEdit model [2009] and our GenSSS model at comparable data sizes. The scenes were rendered at 16 samples / pixel. From top to bottom: blue wax, chessboard (4×4), jade and yellow wax materials on dragon, buddha, dragon and statue objects, respectively:

Reference SubEdit model [2009] Our GenSSS model, <em>K</em>=10
Reference
SubEdit model [2009]
Our GenSSS model, K=10
Reference SubEdit model [2009] Our GenSSS model, <em>K</em>=10
Reference
SubEdit model [2009]
Our GenSSS model, K=10
Reference SubEdit model [2009] Our GenSSS model, <em>K</em>=10
Reference
SubEdit model [2009]
Our GenSSS model, K=10
Reference SubEdit model [2009] Our GenSSS model, <em>K</em>=10
Reference
SubEdit model [2009]
Our GenSSS model, K=10

The visual comparisons between the NMF based model (similar to Peers et al. [2006]) and our GenSSS model at comparable data sizes. The scenes were rendered at 16 samples / pixel. From top to bottom: blue wax, chessboard (4×4), chessboard (8×8) and yellow wax materials on buddha, statue, dragon and lucy objects, respectively:

Reference NMF based model Our GenSSS model, <em>K</em>=5
Reference
NMF based model
Our GenSSS model, K=5
Reference NMF based model Our GenSSS model, <em>K</em>=5
Reference
NMF based model
Our GenSSS model, K=5
Reference NMF based model Our GenSSS model, <em>K</em>=5
Reference
NMF based model
Our GenSSS model, K=5
Reference NMF based model Our GenSSS model, <em>K</em>=5
Reference
NMF based model
Our GenSSS model, K=5

Comparison of difference images

The visual comparisons of the color-coded difference images between the SubEdit model [2009] and our GenSSS model at comparable data sizes. From top to bottom: blue wax, chessboard (4×4), jade and yellow wax materials on dragon, buddha, dragon and statue objects, respectively:

SubEdit model [2009] Our GenSSS model, <em>K</em>=10
SubEdit model [2009]
Our GenSSS model, K=10
SubEdit model [2009] Our GenSSS model, <em>K</em>=10
SubEdit model [2009]
Our GenSSS model, K=10
SubEdit model [2009] Our GenSSS model, <em>K</em>=10
SubEdit model [2009]
Our GenSSS model, K=10
SubEdit model [2009] Our GenSSS model, <em>K</em>=10
SubEdit model [2009]
Our GenSSS model, K=10

The visual comparisons of the color-coded difference images between the NMF based model (similar to Peers et al. [2006]) and our GenSSS model at comparable data sizes. From top to bottom: blue wax, chessboard (4×4), chessboard (8×8) and yellow wax materials on buddha, statue, dragon and lucy objects, respectively:

NMF based model Our GenSSS model, <em>K</em>=5
NMF based model
Our GenSSS model, K=5
NMF based model Our GenSSS model, <em>K</em>=5
NMF based model
Our GenSSS model, K=5
NMF based model Our GenSSS model, <em>K</em>=5
NMF based model
Our GenSSS model, K=5
NMF based model Our GenSSS model, <em>K</em>=5
NMF based model
Our GenSSS model, K=5

Comparison of response images

Side-by-side comparison between the SubEdit model [2009] and our GenSSS model at comparable data sizes:

Response image comparison Response image comparison Response image comparison Response image comparison

Side-by-side comparison between Peers et al.'s model [2006] and our GenSSS model at comparable data sizes:

Response image comparison Response image comparison

Comparison of albedo images

The visual comparisons between the SubEdit model [2009] and our GenSSS model at comparable data sizes. From top to bottom: chessboard (4×4) and jade materials:

Reference SubEdit model [2009] Our GenSSS model, <em>K</em>=10
Reference
SubEdit model [2009]
Our GenSSS model, K=10
Reference SubEdit model [2009] Our GenSSS model, <em>K</em>=10
Reference
SubEdit model [2009]
Our GenSSS model, K=10

The visual comparisons of the color-coded difference images between the SubEdit model [2009] and our GenSSS model at comparable data sizes. From top to bottom: chessboard (4×4) and jade materials:

SubEdit model [2009] Our GenSSS model, <em>K</em>=10
SubEdit model [2009]
Our GenSSS model, K=10
SubEdit model [2009] Our GenSSS model, <em>K</em>=10
SubEdit model [2009]
Our GenSSS model, K=10

Application

Our application can be used to convert heterogeneous translucent materials into homogeneous ones by simply using a rank-1 approximation of SVD. From top to bottom: artificial stone, blue wax, chessboard (4×4), chessboard (8×8), jade and yellow wax materials on dragon, lucy, statue, buddha, dragon and statue objects, respectively:

Reference Our homogeneous model, <em>K</em>=1 Our heterogeneous model, <em>K</em>=5
Reference
Our homogeneous model, K=1
Our heterogeneous model, K=5
Reference Our homogeneous model, <em>K</em>=1 Our heterogeneous model, <em>K</em>=5
Reference
Our homogeneous model, K=1
Our heterogeneous model, K=5
Reference Our homogeneous model, <em>K</em>=1 Our heterogeneous model, <em>K</em>=5
Reference
Our homogeneous model, K=1
Our heterogeneous model, K=5
Reference Our homogeneous model, <em>K</em>=1 Our heterogeneous model, <em>K</em>=5
Reference
Our homogeneous model, K=1
Our heterogeneous model, K=5
Reference Our homogeneous model, <em>K</em>=1 Our heterogeneous model, <em>K</em>=5
Reference
Our homogeneous model, K=1
Our heterogeneous model, K=5
Reference Our homogeneous model, <em>K</em>=1 Our heterogeneous model, <em>K</em>=5
Reference
Our homogeneous model, K=1
Our heterogeneous model, K=5