Weekly Update for February 12

Sample Slide

• The
• groundhog
• was
• right
• :’(

Shoe matching on edges

Shoe matching on edges

Shoe matching on edges

Shoe matching on edges

Shoe matching on edges

Shoe matching on edges

Shoe matching on edges

Shoe matching on edges

Shoe matching on edges

Shoe matching on edges

Shoe matching on edges

Shoe matching on edges

Shoe matching on edges

Shoeprint registration update

• What is working?
  • Given two shoeprint images, we can do a initial rough alignment using an axis found with SLR
  • After this is done, we apply the MATLAB function ‘imregister’ to get a better alignment

Currently working on…

• Writing my version of ‘imregister’ in Rcpp
  • The algorithm is based on minimizing the difference of the pixel intensities after a given affine transformation is applied
  • The optimization algorithm is based on gradient descent
  • The more interesting part is the use of the pyramid representation

Coding things up

• Just coding things up. Thinning algorithm can take up to a minute for full documents
• When we start processing documents, this adds up. Would like to decrease this time
• Converting R code to Rcpp