Journal Home Page

Cumulative Index

List of all Volumes

Complete Contents
of this Volume

Previous Article

Next Article
 


Journal for Geometry and Graphics 21 (2017), No. 1, 131--139
Copyright Heldermann Verlag 2017



Low-Poly Image Stylization

Thitiwudh Uasmith
Dept. of Media Technology, King Mongkut's University of Technology Thonburi, 126 Pracha Uthit Road, Bangkok 10140, Thailand
tdy2020@gmail.com

Tantikorn Pukkaman
Dept. of Media Technology, King Mongkut's University of Technology Thonburi, 126 Pracha Uthit Road, Bangkok 10140, Thailand
pukkarmoo@gmail.com

Peeraya Sripian
Dept. of Media Technology, King Mongkut's University of Technology Thonburi, 126 Pracha Uthit Road, Bangkok 10140, Thailand
midorip@gmail.com



A low-poly image is a minimalist style of art that is currently widely used. It is an image derived from low-polygon 3D objects with an idea of image non-photorealistic abstraction. The trend of using low-poly images has accelerated rapidly since the introduction of vector images. Because low-poly images are based on vectors, the images are compact, scalable, editable, and easy to animate. Although semi-automatic low-poly image conversion tools exist, the resulting graphic generated from the tool does not resemble the original image in terms of its global structure. This paper presents an application that automatically creates a low-poly image that preserves global details, such as edges. Given a raster image, our algorithm automatically computes the polygons that best approximate the image. Our proposed algorithm is mainly based on K-means segmentation, contour extraction, the Ramer-Douglas-Peucker simplification, and the Delaunay triangulation. The output image is in a vector file format that can be easily further manipulated. For evaluation purposes, we distributed questionnaires to 30 participants that were comprised of 30 low-poly image sets; each set contained a low-poly image generated using our method and one generated using another method. From the experiment, we found that the majority of participants preferred the low-poly images generated using our method.

Keywords: Image vectorization, image abstraction, image triangulation, low-poly image.

MSC: 68U05; 68U10

[ Fulltext-pdf  (3411  KB)]