Publications‎ > ‎

Image Completion using Planar Structure Guidance


Our image completion algorithm automatically extracts mid-level constraints (perspective and regularity) and uses them to guide the filling of missing regions in a semantically meaningful way. Our method is capable of completing challenging scenes such as multiple building facades (left), strong perspective distortion (middle) and large regular repetitive structures (right). We significantly outperform three representative state-of-the-art image completion techniques for these images (see Figure 2). Input images (from left to right) are from © Flickr users micromegas, Theen Moy, and Nicu Buculei, used under the Creative Common license. 


Abstract

We propose a method for automatically guiding patch-based image completion using mid-level structural cues. Our method first estimates planar projection parameters, softly segments the known region into planes, and discovers translational regularity within these planes. This information is then converted into soft constraints for the low-level completion algorithm by defining prior probabilities for patch offsets and transformations. Our method handles multiple planes, and in the absence of any detected planes falls back to a baseline fronto-parallel image completion algorithm. We validate our technique through extensive comparisons with state-of-the-art algorithms on a variety of scenes.

Citation

Jia-Bin Huang, Sing Bing Kang, Narendra Ahuja, and Johannes Kopf, Image Completion using Planar Structure Guidance (Proceedings of SIGGRAPH 2014), 33(4), To Appear


Bibtex

@article{Huang-TOG-2014,

author = {Jia-Bin HuangSing Bing KangNarendra Ahuja, and Johannes Kopf}, title = {Image Completion using Planar Structure Guidance}, journal = {ACM Transactions on Graphics (Proceedings of SIGGRAPH 2014)}, year = {2014}, volume = {33}, number = {4}, pages = {to appear}
}


 
Paper 
High-res [PDF] (21.9 MB)
Low-res  [PDF] (11.7 MB)
Official paper [Link]

Supplementary Material
- Man-made 
[Link
- Natural [Link]

Download dataset

- Data 
[Link] (712 MB)
- Readme [Link]

Slides
-
PPT [Link] (160 MB)
- PDF [Link] (8.2 MB)
- Talk [Video]
 https://github.com/jbhuang0604/StructCompletion


Reference code

- [GitHub]


Figure 2: Limitations of current state-of-the-art methods. Compare these results with ours in Figure 1.



More Results

More Results


Sample comparisons with previous works 



   
  
  
  
  
  
  
  

    

          

Failure cases
  
   

Visualization of the completion process in the coarse-to-fine fashion

 
Image credit: ©Flickr user Nicu Buculei
 
Image credit: ©Flickr users aigle_dore
Image credit: ©Flickr user cedrennes 
Image credit: ©Flickr user the_dugghouse
Image credit: ©Flickr user sunshinepictures

Tracker to track web visitors

Comments