摘要： 探讨Lab颜色空间内基于颜色的图像检索问题。分析已有的基于Lab色度直方图的检索算法的不足，提出改进的基于Lab空间颜色通道的检索算法，该算法分别将a, b颜色通道非均匀量化成22个等级，用其直方图来表征图像的颜色特征，从而保留了两个颜色通道的特性。设计并实现了基于用户感兴趣图像块的相关反馈技术。实验结果表明，改进算法具有良好的检索结果，采用的相关反馈技术提高了检索性能。

关键词: Lab空间, 色度值, 颜色通道, 感兴趣分块

Abstract: This paper discusses color-based image retrieval technique in the Lab color space. It analyzes the shortcoming of existing algorithm based on Lab hue histogram, presents an improved retrieval algorithm based on chromaticity coordinate of Lab space. A uniform quantization of chromaticity coordinates(a, b) down to twenty-two bins is performed, their histograms can describle the color context of the image, and preserves the chromaticity coordinates’ characteristic. This paper implements the relevant feedback technique with piece-of-interest. Experimental results show that the improved method has sound and robust retrieval performance, and the relevant feedback technique also enhances the retrieval effectiveness.

Key words: Lab space, hue, color channel, piece-of-interest

中图分类号: 

• TP391.41