基于小波分析的图像检索技术

随着大量数字图像数据库的出现,基于内容的图像检索技术成为研究热点。本文针对小波的多尺度分析特性,描述了一种基于小波分析的图像检索算法,在小波变换域内提取出图像的显著特征点,然后提取显著特征点的颜色和纹理特征,颜色选取HSV空间,纹理特征用Gabor小波幅值的均值与方差表示,通过相似性匹配实现图像的检索。最后,本文通过实验证实了这种方法具有较高的检索精度,与基于角点检测方法的图像检索相比,检索效果与人的视觉更加相符。

Image retrieval technique based on wavelet analysis

With the appearance of a large of digital image databases,content-based image retrieval technology become a popular research field.In this paper,aiming at multi-scale analysis feature of wavelet,we present a wavelet analysis- based image retrieval arithmetic.We first extract salient feature points in wavelet transform domain.Then extracting colour and texture features from these salient feature points to compute the similarity between images,and realize the image retrieval process.HSV space is used for the colour feature,the mean values and standard deviations of Gabor range are used for the texture feature.In this experiment,our method provides better retrieval performance and result comparing with classic corner detectors.