expressive painterly rendering through image processing

Expressive Painterly Rendering Through Image Processing

by Jason Douglas Waltman

A thesis submitted to the faculty of
The University of Utah
in partial fulfillment of the requirements for the degree of

Master of Science in Computer Science

School of Computing
The University of Utah
May 2004

Abstract

In real paintings, a human is in control of every aspect of the painting creation process and, as a result, these works of art are seemingly more valued than their photographic equivalents. The human painter is able not only to document a scene but also to add expressive elements—for example, a stylized abstraction combined with exaggerated brushwork—not possible in a photograph of the same scene. Automatic, computer-generated, photograph-to-painterly rendering techniques have been published now for over a decade. In general, these techniques produce images that lack characteristics found in paintings created by human hands. A new painterly rendering technique is presented which employs digital image processing, computer vision, and emulation of details found in works by painting masters in order to produce images with human expressive characteristics and important details found in real paintings.

Download

An online version of the thesis is not available, however two versions of the PDF may be downloaded here. The low-res version is suitable for reading, but you'll need the 22 MB version to see really good images (results are presented below if you can't wait on the long download).

Waltman Thesis Final 040325.pdf
21.5 MB

Waltman Thesis Final 040325 low-res.pdf
1 MB

Results

Presented here are the more successful results from this work; both the original photograph and the final painterly rendering are shown. As fine detail is important in these renderings, clicking on any image below will download a larger version for your viewing pleasure (at about 75% of the original rendering resolution).

Aspen Trees Original Photograph
Click to enlarge.