Multi-Class Blue Noise Sampling Li-Yi Wei Microsoft Research
smart↑ → research↑ Q: not so smart?
Blue noise random & uniform sample spacing r
r = 0.02 r = single classmulti class
Multi-class blue noise class 0class 1total set
Object placement
Sensor layout cone/rod cellsRGB sensors
Color stippling RGBCMYB dots
A: smart↓ → popular↑ simpler algorithms less intimidating 5456 too smart # sig10 paper
50 seconds old version
smart↑ → research↑ Q: not so smart?
Blue noise random & uniform sample spacing r
r = 0.02 r = single classmulti class
Multi-class blue noise class 0class 1total set
Object placement
Sensor layout cone/rod cellsRGB sensors
Color stippling RGBCMYB dots
A: smart↓ → popular↑ simpler algorithms less intimidating 5456 too smart # sig10 paper
Old Version continue joke from FF 2008
Blue noise distribution random & uniform dart throwing Lloyd relaxation
X not novel [fast-forward SIG 2008] X not very useful texture synthesis inverse texture synthesis flip
overlay class 0class 1 X not uniform total set
Multi-class blue noise class 0class 1total set
Object placement
Sensor layout cone/rod cellsRGB sensors
Color stippling RGBCMYB dots
Backup
overlay X not uniform
Sensor layout continuous domaindiscrete domain retina
texture synthesis
inverse texture synthesis
Old slide from SIGGRAPH 2000 input (small) output (large) texture synthesis
New SIGGRAPH 2008 paper (just by flipping the old slide) input (large) output (small) inverse texture synthesis
Object placement Uniform distribution Red flower Yellow flower Entire set
Uniform per class class 0 Poisson disk class 1 Poisson disk total set OOX
Uniform total set total set Poisson disk class 0class 1 XXO
Our method class 0 Poisson disk class 1 Poisson disk total set Poisson disk OOO