Painterly: Painting-Like Renderings from Portrait Photographs

The painterly project consists of the creation of a toolkit for 'knowledge based' NPR painterly renderings from portrait photographs based on Prof. Steve DiPaola's research at the Intelligent Visualization Lab (http://ivizlab.sfu.ca/).   It uses typical photographs of people as input, along with an XML based script file that specifies a number of parameters based on cognitive knowledge of the open methodology that human portrait artist use. Unlike filter based systems, and more like human painters, the toolkit is able to use knowledge of a portrait sense as well as exploit human vision and perception traits to, as human artists do, "filter out the unimportant, while emphasizing the essence of the scene". This leads to more automatically adaptable, organic and wider NPR results than standard systems. It has applications for computational photography, automatically generated artistic looking games and movies as well as being used as a interdisciplinary toolkit for art scholars, cognitive scientists and computer scientists. Ongoing visual results can be seen at http://ivizlab.sfu.ca/research/painterly.

See http:://painterly.costar.sfu.ca/ for the project source code repository.