Carleton University - School of Computer Science Honours Project
Recolouring Photographic Image with a Palette Set
The focus of this project was to explore recolouring a photographic image using a palette set. This objective was achieved by using a k-means clustering algorithm to determine the palette. After the palette was generated, a distance field was created based on the palette and the image plane. The distances field consisted of all geodesic distances for each palette colour to every pixel. Converting the distances to weights, the image was be recoloured by only manipulating the hue of the original pixels. While only being able to change the hues, pixels that were near to black or white were problematic. Pixels that were close to white and with some saturation value, change to unnatural colours. In addition, only being able to change hues, black pixels were extremely difficult to alter.