Poisson Image Editing: Siggraph 2003 Patric Perez Michel Gangnet Andrew Black
Poisson Image Editing: Siggraph 2003 Patric Perez Michel Gangnet Andrew Black
Poisson Image Editing: Siggraph 2003 Patric Perez Michel Gangnet Andrew Black
Heat Equation
Laplaces Equation
Boundary Conditions
Dirichlet Boundary Conditions
Specify the value of the function on a surface
Guided Interpolation
Simple Interpolation
Maximize the Smoothness
Guided Interpolation
Interpolation-> minimization
Partial Derivative
Discretize
Texture
Alignment
Discretization
Texture Flattening
Remain Only Salient Gradients
Discretization
Texture Flattening
Edge mask
Seamless Tiling
Select original image as g Boundary condition:
f*north=f*south=0.5(gnorth+gsouth) Similarly for the east and west
Seamless Tiling
Discussion
Discussion
Discussion
Discussion
Drag-and-Drop Pasting
SIGGRAPH 2006 Leo Jiaya Jia Jian Sun Chi-Keung Tang Heung-Yeung Shum The Chinese University of Hong Kong Microsoft Research Asia The H.K. University of Sci. & Tech. Microsoft Research Asia
+
fs
ft
+
fs
ft
min f
p0
| f f s |2 dp with f |0 = ft |0
'
Taking f = f f s, we have
min f '
p0
ft
0
Image blending should take both the source and the target images into consideration. Property of solving the Laplace equation:
| f ' |2 will approach zero if and only if all The variational energy 0
min
(( f t ( p ) f s ( p )) k ) 2 , s.t. blue
Given new k, optimize the boundary . Repeat the previous two steps until convergence.
Boundary optimization path between any two In 2D graph, computing the shortest
points: Dynamic Programming Our problem is to compute a closed path
Boundary optimization
Boundary optimization
Boundary optimization
Boundary optimization
One example
Matting
Composition Image I is generated by foreground and background with alpha matte I = alpha*F + (1-alpha)*B Matting is a problem to get alpha,F,B from a given image I. User have to devide the image into three region: Foreground, Background and Unknown area. In foreground area, F = I, alpha = 1, B = 0; In background area, F = 0, alpha = 0, B = I. Our task is to get F,B,alpha in unknown area.
p*
| f v ' |2 dp with f |* = ft |*
Alpha blending
Our method
Poisson blending
Our method
Alpha blending
Our method
Poisson blending
SIGGRAPH 2006 sketch Daniel Leventhal Brown Univ. Bernard Gordon Brown Univ. Peter G. Sibley Brown Univ.
Alpha control
Alpha in [0, 1], background 0, and foreground 1. Obtain unselected areas Alpha by blurring.
Result
Luminance rescaling
Mesh merging
Deformation
Basic idea
Quite similar to the case of image Vector field decomposition can be extended to manifold
Discretization
NOT real gradient and divergence, but it works
Mesh Merging
Some ugly details
Boundary interaction Boundary correspondence Re-parameterization
Mesh deformation
Interactively change guided vector field
Change normal on a curve Propagate to other areas
SPM 2007 Xiaohuang Huang Zhejiang Univ. & HUST Hongbu Fu HUST Oscar Kin-Chung Au HUST Chiew-Lan Tai HUST
Basic idea
Best boundary for Poisson Merging Similar to Drag-and-Drop Pasting.
Optimal boundary
E (, k ) =
p
Results
Failure?