Abstract
Salient object detection techniques have a variety of applications of broad interest. However, the detection must be fast to facilitate these processes. In this paper, we address the computational problems in salient object detection. Several approaches to resolving the salient object detection problem consist of two steps: saliency map extraction and salient object localization. To achieve accurate detection, multiple features are typically combined for computing a saliency map, and a dense-sampling approach that examines numerous regions is widely used (both processes are computationally demanding). We integrated salient feature computation into the search process and accelerated state-of-the-art approaches by using an efficient subwindow search framework. We developed a fast and accurate salient object detection system. The experimental results using the MSRA salient object database validated the effectiveness and the computational efficiency of the proposed approach.
Original language | English |
---|---|
Pages (from-to) | 60-66 |
Number of pages | 7 |
Journal | Pattern Recognition Letters |
Volume | 46 |
DOIs | |
Publication status | Published - 2014 Sept 1 |
Keywords
- Computational efficiency
- Object detection
- Saliency
ASJC Scopus subject areas
- Software
- Signal Processing
- Computer Vision and Pattern Recognition
- Artificial Intelligence