- Geometric registration
- Illumination handling
- Occlusion handling
These topics are covered in detail in my PhD thesis.
Let's imagine that we want to augment this T-shirt with a platipus.
Model image | Virtual character |
The result would look like this:
We can establish wide-baseline keypoint matches between the model image and the input image, and use our geometric registration algorithm to know where to render the virtual character.
can't load bibtex database