ComPhoto
Computational Photography, Fall 2011

Jump to...

project #1
final project
assignments


project #2: Image Blending

Assigned: 2011/11/17
Due: 2011/12/07 11:59pm

Project description

Image blending composites a selected image region into a target image seamlessly. In this project, you are asked to implement two image blending methods: Poisson blending [1] and Coordinate instant cloning [2]. For the first method, you will need to implement a Poisson solver. You can find example codes for Poisson solver from this website. You might need to modify them for your project though. For the second method, you only need to implement the baseline method without speeding up using adaptive meshes. You can work in a team of two students for this project.



Bonus

There are several options for bonus. First, you can speed up coordinate instant cloning using adaptive meshes as described in the paper [2]. Second, you can find the best seam before blending. In the following figure, the seam on the left is speficied by the boundary of the selected region and the seam on the right is a better contour found by the algorithm. The seam can be found by either graph cut (interactive digital photomontage [3]) or shorest closed path (drag and drop pasting [4]). You are free to add proper user inputs for this purpose, such as serveral strokes.



Submission

You need to turn in your source code and a html report which describes what you have done and shows results.
Instructions for submission will be posted later. Please submit your project through this website. If you are within cmlab domain, use this one instead.

Reference

  1. P. Perez, M. Gangnet, Poisson Image Editing, ACM SIGGAPH 2003.
  2. Z. Farbman, G. Hoffer, Y. Lipman, D. Cohen-Or, D. Lischinski, Coordinates for Instant Image Cloning, ACM SIGGAPH 2009.
  3. A. Agarwala, M. Dontcheva, M. Agrawala, S. Drucker, A. Colburn, B. Curless, D. Salesin, M. Cohen, Interactive Digital Photomontage, ACM SIGGAPH 2004.
  4. J. Jia, J. Sun, C.-K. Tang, H.-Y. Shum, Drag and Drop Pasting, ACM SIGGAPH 2006.