Vectorization involves turning a pixelated raster image into a series of smooth mathematical curves. It has applications in animation, industrial design, and illustration. Vectorization faces many challenges, including noise, ambiguous junctions, and superfluous details; even commercial software leaves much to be desired in terms of accuracy. This project examines several state-of-the-art vectorization techniques and uses them to implement a simple vectorization program. It also assesses the strength & weaknesses of its algorithms and provides concrete suggestions on how the software could be improved, both in terms of robustness and accuracy. While the resulting vectorized images lack in detail, the software nonetheless manages to capture the overall outline of each image. In addition, several original improvements to the algorithms are described, including a solution to address filled areas, the implementation of a minimum speed factor, and the possibility of using corners to improve centreline extraction.