Upload
keenan
View
37
Download
0
Embed Size (px)
DESCRIPTION
Image Repairing: Robust Image Synthesis by Adaptive N D Tensor Voting. IEEE Computer Society Conference on Computer Vision and Pattern Recognition. Jiaya Jia, Chi-Keung Tang Computer Science Department The Hong Kong University of Science and Technology. Motivation. - PowerPoint PPT Presentation
Citation preview
Image Repairing: Robust Image Synthesis by Adaptive ND Tensor Voting
IEEE Computer Society Conference on Computer Vision and Pattern Recognition
Jiaya Jia, Chi-Keung TangJiaya Jia, Chi-Keung Tang
Computer Science DepartmentComputer Science DepartmentThe Hong Kong University of The Hong Kong University of
Science and TechnologyScience and Technology
Motivation
• Main difficulties to repair a severely damaged image of natural scene– Mixture of texture and colors– Inhomogeneity of patterns– Regular object shapes
Motivation
• Given as few as one image without additional knowledge, we address:– How much color and shape information in the
existing part is needed to seamlessly fill the hole?– How good can we achieve in order to reduce
possible visual artifact when the information available is not sufficient.
• Robust Tensor Voting method is adopted
Tensor Voting Review• Tensors: compact representation of information • Tensor encoding:
3D tensor
3
1 2Ball tensor: uncertainty
in all directions
Plate tensor: certainty of directions in a plate
Stick tensor: certainty along two opposite directions
Tensor Voting Review
• Voting process is to propagate local information
P
Osculating circle
Image repairing system
Input Damaged Image
Texture-based Segmentation
Statistical Region Merging
Curve Connection
Adaptive Scale Selection
NND Tensor Voting
Output Repaired Image
Complete Segmentation
Image synthesis
SegmentationSegmentation
• JSEG [Deng and Manjunath 2001] – color quantization – spatial segmentation
• Mean shift [Comanicu and Meer 2002]
• Deterministic Annealing Framework [Hofmann et al 1998]
Texture-based SegmentationTexture-based Segmentation
Statistical Region Merge
• (M + 1)D intensity vector for each region Pi,
where M is the maximum color depth in the whole image.
0
20
40
60
80
100
1 2 … M M+1
PiPk
histogram gradient
if
1MiV
( 1) || ||i
ij Pi
V M jN
i kP P P
, || ||i k i ks V V Threshold
Why Region Merge?
• Decrease the complexity of region topology
• Relate separate regions
P1
P5
P3 P4
Damaged area
P2
Curve Connection
• 2D tensor voting method
P1
P5
P3 P4
P2
Z
XP2 P4
Why Tensor Voting?
• The parameter of the voting field can be used to control the smoothness of the resulting curve.
• Adaptive to various hole shapes
Small ScaleSmall Scale
Large ScaleLarge Scale
Without hole constraint
Without hole constraint
With hole constraint
With hole constraint
P4
Connection Sequence• Topology of surrounding area of the hole can be
very complex• Greedy algorithm
– Always connect the most similar regions
P1
P5
P3
Damaged area
P2
P2 and P4
P3 and P5
P1
Complete Segmentation
Image repairing system
Input Damaged Image
Texture-based Segmentation
Statistical Region Merging
Curve Connection
Adaptive Scale Selection
NND Tensor Voting
Output Repaired Image
Complete Segmentation
Image synthesis
ND Tensor Voting• Tensor encoding
– Each pixel is encoded as a ND stick tensor
5
5
Stick tensorScale N=26
ND Tensor Voting
• Voting process in ND space– An osculating circle becomes an osculating
hypersphere.– ND stick voting field is uniform sampling of normal
directions in the ND space.
sample sample
Adaptive Scaling
• texture inhomogeneity in images gives difficulty to assign only one global scale N [Lindeberg et al 1996].
• For each pixel i in images, we calculate:
{( )( ) }i i
TN N
M AVG I I {( )( ) }i i
N N
TM AVG I I {( )( ) }i i
TN N
M IAVG I • trace(M) measures the
average strength of the square of the gradient magnitude in the window of size Ni
Adaptive Scaling
• For each sample seed:– Increase its scale Ni from the lower bound to the
upper bound– If trace( ) < trace( ) - α where α is a
threshold to avoid small perturbation or noise interference, set Ni - 1 → Ni and return
– Otherwise, continue the loop until maxima or upper bound is reached
iNM 1iNM
Results
Results
Results
Results
Results
Limitations
• Lack of samples.
• Meaningful and semi-regular objects.
Conclusion
• An automatic image repairing system.
• Region partition and merging.
• Curve connection by 2D tensor voting.
• ND tensor voting based image synthesis.
• Adaptive scale.