综合视觉注意模型的显著性局部特征提取算法研究

Salient Local Feature Extraction Algorithm Based on Integrated Visual Attention Model

图像检索过程中往往会提取大量的局部特征,这将加大图像检索的计算量和复杂度,影响其应用。针对这一问题,提出了一种应用综合视觉注意模型的显著性分析提取局部特征的方法:在图像尺度空间中提取关键点,利用模糊增长技术查找原始图像的显著性区域,计算其综合视觉显著性权值并分类,提取SIFT描述因子,保留最突出的局部特征以提高检索性能。相比于传统的局部特征提取算法,本方法在图像检索精度和检索速度方面都具有明显优势。

Local features are widely used for content-based image retrieval recently. During image retrieval,a lot of local features are extracted,which increases the amount of calculation and complexity of image retrieval, and as a result, af- fecting the practical applications. With an eye towards this problem, a novel method based on integrated visual attention model was proposed to extract salient local features. Using this method, first, the key points in an image scale-space are extracted,and the salient area of the original image is found using fuzzy growth technology, then the integrated visual saliency is calculated and classified,and SIFT factors are extracted and ranked according to their integrated visual sali- ency, and at last,only the most distinctive features are kept to enhance the retrieval performance. The experimental re- sults demonstrate that compared to traditional local feature extraction algorithms, this salient local feature extraction al- gorithm based on integrated visual attention model provides significant benefits both in retrieval accuracy and speed.