中文摘要：

相似度量是图像检索的关键，EMD是一种有效的度量距离，但其计算比较复杂，而且依赖于基本距离的选择。采用Lloyd聚类算法对图像进行高斯混合建模，并以聚类失真作为基本距离，提出了两种近似EMD的方法计算相似度。实验结果验证了该方法的有效性，其检索效率与EMD方法接近，而且计算复杂度比EMD方法低，基本距离的选择不敏感。

英文摘要：

Similarity measure is crucial for content based image retrieval, EMD is an efficient distance measurement, but it costs high computation complexity, and relies on efficient ground distance. In this paper, the images are modeled with Lloyd clus- tering, and two methods to approximate EMD with clustering distortion as ground distance are proposed. Experiment results demonstrate the effectiveness of the pro-posed methods. With similar performance compared with EMD, the proposed methods not only have lower computation complexity, but aren＇t sensitive to the ground distance.