A two-stage region-based image retrieval approach using combined color and texture features

Authors:
Yinghua Lu;Qiushi Zhao;Jun Kong;Changhua Tang;Yanwen Li
Affiliations:
Computer School, Northeast Normal University, Changchun, Jilin Province, China;Computer School, Northeast Normal University, Changchun, Jilin Province, China;Computer School, Northeast Normal University, Changchun, Jilin Province, China;Computer School, Northeast Normal University, Changchun, Jilin Province, China;Computer School, Northeast Normal University, Changchun, Jilin Province, China
Venue:
AI'06 Proceedings of the 19th Australian joint conference on Artificial Intelligence: advances in Artificial Intelligence
Year:
2006

Citing 8
Cited 0

Color indexing

International Journal of Computer Vision
Color matching for image retrieval

Pattern Recognition Letters
Efficient Color Histogram Indexing for Quadratic Form Distance Functions

IEEE Transactions on Pattern Analysis and Machine Intelligence
An effective region-based image retrieval framework

Proceedings of the tenth ACM international conference on Multimedia
Blobworld: A System for Region-Based Image Indexing and Retrieval

VISUAL '99 Proceedings of the Third International Conference on Visual Information and Information Systems
Image retrieval system based on color-complexity and color-spatial features

Journal of Systems and Software
A novel fusion approach to content-based image retrieval

Pattern Recognition
Frequency layered color indexing for content-based image retrieval

IEEE Transactions on Image Processing

Quantified Score

Hi-index	0.00

Visualization

Abstract

An effective Content-Based Image Retrieval (CBIR) approach is proposed in this paper. In contrast with existing systems, the retrieval process is divided into two stages. Images are firstly classified into categories based on their semi-global features automatically using the Fuzzy C-Means (FCM) clustering algorithm, and the K Nearest Neighbor (KNN) algorithm is used to assign the query image into a proper category to get a candidate image set. As a consequence, most irrelevant images are pruned. For the second stage, a novel segmentation algorithm is applied to segment both the query image and the candidate images into regions approximately according to objects. Color and texture features are extracted from each region for finer level retrieval. The region-based features utilize local properties of objects in image, and it is suitable for complicated scenes. Finally, distance measure is applied to evaluate the image-level similarity. This coarse-to-fine mechanism provides an effective and efficient performance for our system, which is demonstrated in the experiments on the image database from COREL.