基于改进的受限随机选择的图像检索相关反馈  

RELEVANCE FEEDBACK IMAGE RETRIEVAL BASED ON MODIFIED CONSTRAINED RANDOM SELECTION

作　　者：张海洋[1] 

机构地区：[1]宿州学院地理与环境科学系,安徽宿州234000

出　　处：《巢湖学院学报》2008年第3期69-72,共4页Journal of Chaohu University

摘　　要：文章提出了一种基于改进的受限随机选择算法的图像检索相关反馈。首先通过移动查询点更新查询向量,然后针对受限随机方法中的等距划分未能反映负样本真实分布的缺点进行改进,提出根据等密度划分来确定负样本选择的子区域。改进方法提高了负样本的代表性,并且一定程度上解决了反馈中的"小样本"问题,实验证明,较原有的受限随机选择能在更少地反馈次数内得到更好地反馈结果。A modified constrained random selection approach for relevance feedback is proposed in this paper. Firstly, update retrieval vector using query point movement algorithm. Then, determine sub-region for negative samples chosen by proposed equal density dividing, since the equal distance dividing methods in the original constrained random selection cannot reflect the real data distribution. The modified method improves the representative of the negative samples, and solves the ＂small size sample problem＂ at certain degrees. Experiments demonstrated that better feedback results have got within less feedback times, comparing with original constrained random selection approach.

关 键 词：移动查询点 受限随机选择 等密度划分 图像检索 相关反馈 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]