点击次数：

论文类型：期刊论文

发表时间：2010-01-01

发表刊物：Journal of Information and Computational Science

收录刊物：EI、Scopus

卷号：7

期号：5

页面范围：1117-1128

ISSN号：15487741

摘要：The proposed system, which absorbs the essence of the saliency map and SIFT algorithm, includes both the online step and the online step. At first, a feature database is established to simulate the human memory in the online operations. Then the saliency region, which is considered as the focus of attention (FOA), will be figured out by the Saliency-Map method in the online operations. Since the SIFT algorithm processes the FOA only, it can reduce the computing time and promote the image-processing efficiency greatly. The holistic model can be regarded as a visual attention model which integrates the bottom-up attention and the top-down attention. The bottom-up component is based on the Saliency-Map method while the top-down component is achieved by the SIFT descriptors. The experimental results show that it takes about 6s to recognize a target from an image with a resolution of 752    480 by using the improved algorithm. Compared with the standard SIFT representation, the proposed model is superior in certain aspects, especially of the efficiency. Copyright ?. 2010 Binary Information Press.