Abstract: An approach is proposed based on image color and texture features combined with genetic algorithm, where in color moments are the color histogram entropy values and the variances of the multiscale high frequency sub-bands of wavelet domain. This method amends the disadvantage of lacking spatial information of common color feature. The more important aspect is that this method cuts down the work of users’ while using relevant feedback by using the selfadaptive character of Genetic Algorithm. By comparative experiments, the experiments show that the approach has good retrieval performance.

Key words: image retrieval, Content Based Image Retrieval（CBIR）, information entropy, multi-scale texture, genetic algorithm, adaptive

摘要： 颜色特征采用分块的HSV颜色空间的信息熵，纹理特征采用小波多尺度高频子带方差特征，结合遗传算法的图像检索方法。采用组合特征进行图像检索，改善了颜色特征缺乏空间信息的缺点，利用遗传算法能够自适应地搜索最优解，减少了在相关反馈的检索过程中，用户的选择操作。通过比较实验，具有很好的检索性能。

关键词: 图像检索, 基于内容的图像检索（CBIR）, 信息熵, 多尺度纹理特征, 遗传算法, 自适应