Home
People
Research
Links
Sponsors
Inpainting is the process of filling the missing
region based on the surrounding image statistics.
Inpainting techniques have been used in
applications ranging from restoration of damaged
old paintings and photographs to removal or
replacement of selected objects in multimedia
editing. Initial efforts are focused on generating
suitable textures by texture synthesis approaches
which generate new textures by sampling and
copying pixels from neighboring areas.
Structural Image Inpainting
These types of "local-feature" based
techniques fail when the hole becomes large.
The reason is that they fail to take advantage
of the many global visual cues like symmetry
and depth for inpainting. For example, our
automatic scheme can exploit the object
symmetry to fill in the miss contour in the left
image. We call this research "Structural
Image Inpainting".
Texture Inpainting Example
Active Research Areas:
Collaborators:
Dr. Ruigang Yang (University of Kentucky)