基于小波包变换的纹理图像检索

Texture Image Retrieval Based-on Wavelet Package Transformation

提出了一种利用小波包变换分解图像,提取纹理特征进行图像检索的新方法。该方法利用小波包函数的特性,对多分辨分析没有细分的高频部分进行分解,从而形成能够精确表征图像的特征矢量。匹配算法则使用比值相似度。实验结果表明,该方法计算简单、有效、检索效果较好。

A new method for image retrieval is presented which decomposes image and extracts image's texture features using wavelet package transformation.Taking advantage of wavelet function,the method can discompose image's high-fre-quency parts so as to form feature vectors that can represent the image accurately,but multi-resolutions analysis can't do it.Besides,the ratio similarity way is adopted in match algorithm.The experimental results have shown that the method improve the retrieval accuracy efficiently and obtain the satisfied retrieval results.