For this project, the shadow removal method used by Finlayson et al. in [1] was implemented.

This report contains an overview of the mathematical background and a detailed

discussion on the experiments performed with the implementation.

This method consists of two independent steps. First, an illumination invariant image

is derived from a color image for which an appropriate color calibration is known. Second,

shadow edges are detected by finding edges in the intensity image that are not in the illumination

invariant image. Gradient information is then erased at those shadow edges, and

a new shadowless color image is formed by solving a Poisson’s equation.