摘要
为加强图像检索领域中图像有效特征的提取,提出一种基于对象提取的图像检索技术。该方法采用二维Otsu梯度阈值选取的快速迭代算法,利用灰度级-最大梯度二维直方图进行阈值选取,同时采用迭代思想代替穷举搜索,快速选取阈值进行对象分割,提取出目标对象,克服了传统二维Otsu算法分割效果不够准确、计算复杂度较高的缺点。将分割出的目标对象进行特征提取与比对,用于检索系统进行检索。实验结果表明,该算法能够较准确的分割出适于图像检索的对象,同时可以得到较高的查准率。
To strengthen the effective feature extraction in the field of image retrieval, an image retrieval technology based on object-extracted is proposed. Two-dimensional Otsu gradient threshold selection fast iterative algorithm is used, the gray level-the maximum gradient two-dimensional histogram is utilized to select threshold, and the iterative is used thought instead of exhaustive search to segment object quickly at the same time. The shortcomings of segmentation inaccurate and the high computational complexity in the traditional two-dimensional Otsu algorithm were overcome. The target object feature is used to extracted and matched for retrieval. Experiments show that the proposed method could segment images more accurate, and the algorithm for object-based image retrieval with higher precision.
出处
《计算机工程与设计》
CSCD
北大核心
2013年第7期2450-2454,共5页
Computer Engineering and Design
基金
国家自然科学基金项目(60775010
61175115)
北京市教委科研计划基金项目(KM201010005012)
