Skip to content

Topic 3: Image inpainting

  1. Read and discuss the paper by Criminisi et al..
  2. Describe the advantages/disadvantages of pixel-based versus patch-based filling approaches.
  3. Investigate the performance of the algorithm for different patch sizes.
  4. Test your implementation on “Kanizsa Triangle”, the synthetic image shown in the paper as well as other test images depicted in the paper.
  5. Discover the limitations of the proposed method using some test images.
  6. Synthesize your own test images and evaluate the performance of the algorithm. As an example, you can capture images of the same scene with the presence and then absence of an object in it. Then try to remove the object (inpaint) and compare it with the ground-truth.

Hint

Usually, you can find the test images of published papers in the supplementary data of the paper online. If not, try to contact the authors to get the original images. If that doesn’t help, try to capture similar images by yourself.

References

A. Criminisi, P. Perez and K. Toyama, “Region filling and object removal by exemplar-based image inpainting,” in IEEE Transactions on Image Processing, vol. 13, no. 9, pp. 1200-1212, Sept. 2004, doi: 10.1109/TIP.2004.833105.

Last update: 2023-04-17