内容图像检索IALA算法的设计与实现

Design and Implementation of IALA Algorithm in Image Retrieval

基于图像检索的距离度量算法主要有两类:一种是最优化方法,其代表算法是旅行商距离(EMD);另一种是统计方法,其代表算法是渐进似然估计(ALA)距离。IALA图像检索算法利用分层的方法来提高检索精度,既克服了最优化方法EMD算法对混合成分发散的源图像的检索效果不佳的缺点,又对传统ALA算法对高斯混合模型方差较大的数据库图像产生误判的不足之处进行了改进。实验证明IALA图像检索算法可以大大提高检索的效率和精度。

There are two categories to the distance measure based on image retrieval algorithm： one is the optimization method,the representative algorithm is the Earth Mover’s Distance（EMD）,the other is the statistical method,the representative algorithm is asymptotic likelihood approximation（ALA） distance.IALA algorithm uses a tiered approach to improve the retrieval accuracy,not only overcomes the shortages in EMD algorithm that divergence of mixed composition for the search result is not satisfactory,but also improves the shortages in traditional ALA algorithm that makes a larger variance Gaussian mixture model for the database image resulting miscarriage of justice.Experiments show that the IALA image retrieval algorithm can greatly improve the retrieval efficiency and accuracy.