一种基于内容图像检索的半监督和主动学习算法

为了提高图像检索中相关反馈算法的效率,提出了一种新的基于相关概率的主动学习算法SVMpr,并结合半监督学习,设计了基于半监督的主动学习图像检索框架。在相关反馈过程中,首先利用半监督学习算法TSVM对标记样本进行训练,然后根据提出的主动学习算法从未标记图像中选取k幅有利于优化学习过程的图像并反馈给用户标记。与传统的相关反馈算法相比,该文提出的图像检索框架显著提高了学习器的效率和性能,并快速收敛于用户的查询概念。

Semi-supervision and Active Relevance Feedback Algorithm for Content-Based Image Retrieval

To improve the efficiency of relevance feedback in image retrieval,a novel active learning algorithm SVMpr based on relevance probability is proposed in the paper.And combined with a semi-supervision learning, the framework of image retrieval based on semi-supervision and active learning is designed.In the process of relevance feedback, firstly, the labeled samples are trained by TSVM,secondly according to active learning algorithm in this paper,the k image samples from the unlabeled images are selected into the user which is added label and more helpful for useful sam-ples.Experimental result shows that the algorithm is better than traditional one,which is in improving significantly the efficiency and convergent rate of machine learning.