Image composition, which is the process of creating a new image by imposing an object or a region from a source image to a target image, is a large component of image editing. Poisson image editing has been proposed in 2003 as an effective approach for seamless image composition.
The effectiveness of Poisson image editing, however, largely depends on how user draws the boundary. Drag and drop pasting in 2006 proposed a method to optimize the boundary condition based on a shortest closed-path algorithm in {r,g,b} color space, and it improved the composition quality by searching for the optimal boundary within the user-specified one.
The algorithm patches up holes in images by finding similar image regions in the database that are not only seamless but also semantically valid.
Initial Project Plan
Mid-term Report
To be released
