小波变换的兴趣特征图像检索技术

Image retrieval based on interested characters of discrete wavelet transform

针对分析过程中优越的多分辨分析和空间频率局部化性能,在离散小波变换(DWT)基础上,提出了一种新的基于内容的图像检索方法.YUV色彩空间中更利于提取颜色特征,将每个图像从标准RGB色彩空间转换到YUV色彩空间,图像的每个分量(Y、U和V)转化为小波域.图像数据库建立阶段,存储每幅图像的小波系数.图像检索阶段,系统比较查询图像与数据库中图像的Y、U和V分量的小波系数,加上用户分配的权重因素,在用户感兴趣特征基础上查询最佳匹配结果.实验结果表明建立方法的是有效的.

Considering the superiority in multiresolution analysis and spatial frequency locallzauon, mls paper propsed a new content-based image retrieval method based on the discrete wavelet transform（DTW）. Because extraction of colour character is simple in YUV colour space, each of the image would be transformed from RGB format to YUV format, and each componet of original image was transformed into wavelet domain. In the image database establishing step, wavelet coefficients of every image were stored. In the image retrieving step, wavelet characters between query image and image in database were compared coupled with the weight factors and optimal results could be obtained. Experimental results show that the method is effective.