Image retrieval based on relevance feedback algorithm by the iteration point will fall into the local optimum. So an adaptive relevance feedback algorithm is proposed, which is able to address such a problem mentioned above. If query point reaches optimum, this query point is automatic splitted into two query points using modified EM algorithm. Two points as original query points, and then reach local optimum respectively. If users do not satisfy accuracy, further more these two points are splitted into four query points, and achieve local optimum. The process of iteration is not stopped until a user is satisfied. Experimental results show that, compared with other algorithms, the algorithm has higher precision rate.%基于相关反馈算法的图像检索经迭代后查询点会陷入局部最优.针对该问题,提出一种基于自适应相关反馈算法的图像检索方法.如果当前查询点达到局部最优,则采用EM裂项算法将该点分解为2个子查询点,通过移动查询点使其各自达到局部最优.如果用户对当前查询不满意,再将这2个子查询点分解为4个子查询点进行处理,以此类推,直到用户满意为止.实验结果表明,与自适应的相关反馈算法、混合反馈算法以及不对称的贝叶斯相关反馈算法相比,该方法的查准率较高.
展开▼
