Abstract: Describing and extracting image’s feature is a key question in content-based image retrieval system，this paper puts forward a new image retrieval method using image information entropy and fractal coding.First，each image in the database is classified in computing information entropy compared with a given threshold.Second，the query image’s fractal coding is got with Jacquin method，which is applied to the same kind of database images as initial image with fractal iteration decoding.Finally，image retrieval result is got by matching the similar distance of the query image and the iterated decoding image.Experimental results show that compared with the direct image pixels similar matching method，our scheme improves the retrieval time greatly and guarantees the retrieval accuracy，thus our proposed method is effective and feasible.

Key words: fractal coding, image information entropy, image retrieval, iterated function system

摘要： 图像的抽象描述和特征提取是基于内容的图像检索系统中需要解决的关键问题，提出了一种图像熵和分形编码相结合的图像检索方法。首先，计算图像熵和比较设定的阈值对图像库进行预分类；其次，利用Jacquin方法计算得到查询图像的分形IFS编码，把图像库同类图像作为初始图像进行分形迭代解码；最后，计算解码图像与查询图像的相似距离得到检索结果。实验结果表明，与直接像素值相似匹配方法相比，在基本保证图像检索效率的前提下，极大地提高了检索时间，该算法具有很好的有效性和可行性。

关键词: 分形编码, 图像熵, 图像检索, 迭代函数系统