Abstract: An image searching algorithm with image indexes is presentde. The first step of this algorithm is to pre-process the image by normalizing it to same size and color space, then divides the image into high frequency and low frequency with 4-layer-wavelet transform. Compute standard deviation in dominant frequency part for the first searching, the second searching is using Euclid distance to measure the similarity with original image, at last, compare the invariant moments to attain the final searching results. In order to verify the effectiveness of the algorithm, we build a framework for it by a searching experiment of 9000 images, the results show that this algorithm is effective.

Key words: Content Base Retrieval, Wavelet Transform, Invariant Moments, Euclid Distance

摘要： 本文主要提出了一个提取图像特征参数来进行多级匹配的检索方法。首先对图像进行预处理，把图像统一转换成相同的尺寸和色彩空间。然后对图像进行4级Daubechies’s 小波变换，并在经过小波变换的图像上进行标准差计算匹配，欧氏距离计算匹配，不变矩计算匹配，最后得出检索结果。本方法通过了有9000多张图像的图像库的检验，获得了很好的实验结果。

关键词: 基于内容的图像检索, 小波变换, 不变矩, 欧氏距离