IFS分形码的快速图像检索算法

Image retrieval algorithm based on IFS fractal code

图像经分形编码后产生IFS分形码,它可被用来进行图像检索操作。针对图像检索的特点,将分形码中的位置参数替换为相对距离与方向系数。定义了分形码间的距离以及图像间的分形码距离,并取出分形码距离最小的前n幅图像作为检索结果,由此提出了基于IFS分形码的快速图像检索算法。从时间复杂性上分析,利用本文算法所需的检索时间与值域块的个数有关。实验结果表明,相对缩放与旋转变化,算法对位移与亮度变化具有较强的稳定性,其分形码距离的均值仅为14.07和20.05;并可检索到具有一定相似性的图像,且类间与类内分形码距离约相差8,类内距离远小于类间距离。

IFS fractal code which is the result of original image fractal coding can be used to retrieve image. In view of feature of image retrieval, the position parameter of IFS code is replaced with the relative distance and direction parameter. The distance between fractal codes and the distance of fractal codes between images are defined, and the preceding n frame images which are the smallest distance sum of fractal code are taken as the retrieval result. Thus, a fast algorithm of image retrieval on IFS code is presented. The retrieval time has a relation to the number of domain blocks by using the proposed algorithm. The experimental results show the algorithm has great stability with the translation and variation of illumination, and the mean of fractal code distance is only 14.07 and 20.05, respectively. The similarity images can be retrieved and the difference between inter-class and inner-class distance is about 8. The inner-class distance is much smaller than the inter-class distance.