Combining Low-Level Features for Semantic Extraction in Image Retrieval
- PDF / 2,145,923 Bytes
- 12 Pages / 600.05 x 792 pts Page_size
- 72 Downloads / 234 Views
Research Article Combining Low-Level Features for Semantic Extraction in Image Retrieval Q. Zhang and E. Izquierdo Multimedia and Vision Laboratory, Electronic Engineering Department, Queen Mary University of London, London E14NS, UK Received 9 September 2006; Revised 28 February 2007; Accepted 16 April 2007 Recommended by Hyoung Joong Kim An object-oriented approach for semantic-based image retrieval is presented. The goal is to identify key patterns of specific objects in the training data and to use them as object signature. Two important aspects of semantic-based image retrieval are considered: retrieval of images containing a given semantic concept and fusion of different low-level features. The proposed approach splits the image into elementary image blocks to obtain block regions close in shape to the objects of interest. A multiobjective optimization technique is used to find a suitable multidescriptor space in which several low-level image primitives can be fused. The visual primitives are combined according to a concept-specific metric, which is learned from representative blocks or training data. The optimal linear combination of single descriptor metrics is estimated by applying the Pareto archived evolution strategy. An empirical assessment of the proposed technique was conducted to validate its performance with natural images. Copyright © 2007 Q. Zhang and E. Izquierdo. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
1.
INTRODUCTION
The problem of retrieving and recognizing patterns in images has been investigated for several decades by the image processing and computer vision research communities. Learning approaches, such as neural networks, kernel machines, statistical, and probabilistic classifiers, can be trained to obtain satisfactory results for very specific applications [1–3]. Unfortunately, fully automatic image recognition using highlevel semantic concepts is still an unfeasible task. Though low-level feature extraction algorithms are well understood and able to capture subtle differences between colors, statistic and deterministic textures, global color layouts, dominant color distributions, and so forth, the link between such lowlevel primitives and high-level semantic concepts remains an open problem. This problem is referred to as “the semantic gap.” To narrow this gap is a challenge that has captured the attention of researchers in computer vision, pattern recognition, image processing, and other related fields, evidencing the difficulty and importance of such technology and the fact that the problem remains unsolved [4, 5]. In this paper, an object-oriented approach for semanticbased image retrieval is presented. Two important aspects of semantic-based image annotation and retrieval are considered: retrieval of images containing a given semantic con-
cept and fusion of different low-level features. The first aspect relate
Data Loading...
