Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
6938120 | Journal of Visual Communication and Image Representation | 2018 | 12 Pages |
Abstract
Color is a rich source of visual information for the effective characterization of image content. The recognition of texture or shape elements in images is strongly associated with the analysis of the image color layout. This paper presents a contextual color descriptor designed especially to be applied to CBIR tasks in heterogeneous image databases. The proposed color uniformity descriptor (CUD) clusters perceptually similar image color regions according to the uniformity analysis of their neighbor pixels. CUD produces vast color image details with a thin histogram, whilst preserving the balance between uniqueness and robustness. CUD is computationally efficient and can achieve high precision and throughput rates when used in CBIR. Experimental results show that CUD performs comparably against local features and multiple features state-of-the-art approaches that require more complex data manipulation. Results demonstrate that CUD provides strong image discrimination even in the presence of significant content variation.
Related Topics
Physical Sciences and Engineering
Computer Science
Computer Vision and Pattern Recognition
Authors
Carolina Reta, Jose A. Cantoral-Ceballos, Ismael Solis-Moreno, Jesus A. Gonzalez, Rogelio Alvarez-Vargas, Nery Delgadillo-Checa,