基于颜色直方图的图像检索技术

Histogram-based color image retrieval

讨论了基于颜色直方图的图像检索中颜色空间和直方图距离的选择问题,提出了利用颜色直方图来检索图像的一般框架,建立了基于两种颜色空间(RGB和HSV)和四种直方图距离共八种图像检索算法,并且采用了两种客观评价标准对所建立算法的有效性进行评价。结果证明,在基于颜色直方图的应用中,HSV空间比RGB空间更有效,EMD距离在所有四种距离量度中性能较好,但计算量稍大,与EMD距离相比,虽交集距离性能稍有不足,但所需计算量较小。

The problem of color space and histogram distance in histogram-based image retrieval is discussed, and the common frame of histogram -based image retrieval is proposed. Then eight kinds of image retrieval algorithms based on two color spaces and four histogram distance algorithms are created and also two objective metrics for evaluating their effectiveness are adopted. The results show that HSV color space is more effective than RGB color space in common. For histogram distance, EMD has showed better performance than other distance, but it has a relative greater computational burden. Comparing with EMD, histogram intersection distance needs less computational burden and only has slightly weak performance.