融合Contourlet和高斯描绘子的图像检索方法

Image retrieval based on Contourlet transformation and Gaussian descriptors

提出一种基于纹理特征和形状特征融合的图像检索方法。采用Contourlet变换提取纹理特征,高斯描绘子提取形状特征,并对两种特征进行融合,最后结合Adaboost算法进行相关性反馈。实验结果表明,所提出的融合方法在图像检索的综合性能上优于Contourlet方法以及基于Gabor小波和高斯描绘子的融合方法,能够有效地实现外观专利设计图像检索。

An image retrieval method is proposed based on fusion features,the texture features are extracted by Contourlet transform and sharp features extracted by Gaussian descriptors.An adaboost algorithm is employed for relevance feedback. The experimental results indicate that the proposed method is superior to the Contourlet method and the method based on Gabor wavelet and Gaussian descriptors,which is available for designing patent image retrieval.