Non-Photorealistic Rendering and Content-Based Image Retrieval

In this paper, we will show how non-photorealistic rendering (NPR) can take a new role in content-based image retrieval (CBIR). We propose a content-based image retrieval method. The novelty is that it is based on the painting representation of images, obtained by Stochastic Paintbrush Transformation (SPT), which automatically simulates a painting process. The painting representation is a stroke sequence. Stroke parameters include color and structure (size, orientation, and position) information. We have compared our method with Global Histogram Intersection (GHI) and Oracle’s CBIR functions on a database of 1,017 images. The results show that our method has a better performance.

• Type: Conference
• Year: 2003
• Where: PG
• Authors: Xiaowen Ji, Zoltan Kato, Zhiyong Huang
• Keywords: Content-based Image Retrieval, Image Retrieval Method, Painting Representation, PG 2003