一种基于相邻匹配的分形图像检索算法  

Fractal Image Retrieval Algorithm Based on Contiguous-matches

作　　者：张思思[1] 刘宇[1,2] 赵志滨[1,3] 

机构地区：[1]东北大学信息科学与工程学院,沈阳110819 [2]河北大学数学与计算机学院,保定07100 [3]东北大学医学影像计算教育部重点实验室,沈阳110819

出　　处：《计算机科学》2015年第12期292-296,311,共6页Computer Science

基　　金：国家自然科学基金(61173027)资助

摘　　要：分形码用来描述图像内跨尺度相似性冗余信息。通过分形码记录图像特征并将其用于图像相似度判断及图像检索。基于自适应四叉树分割方法,提出了图像快速分形编码方法。所提方法通过邻域内固定块的相似性判别快速提取分形码,减少了图像分割层次,缩短了编码时间并保证了图像解码质量。同时提出了一种新的快速判别图像间相似块的距离公式,提升了图像相似度判断的准确性。实验结果表明,相对于灰度直方图判别法,本算法大幅提高了图像检索的查全-查准率。相比于文献中的分形检索算法,本算法缩短了编码时间并降低了分割块数,从而提高了检索效率。Fractal code is used to describe the image similarity in scales redundancy information. In this paper, the image features are recorded as fractal codes that are used to determine the image similarity and retrieval. Based on the self-aaptive method of quadtree segmentation, a new fast fractal image coding method was presented. In this method, fractal codes are gotten quickly from the similarity determination of fixed blocks within neighborhood. The segmentation is decreased and the coding time is reduced with ensuring the quality of image decoding. Meanwhile, the paper also proposed a new distance formula by which the image similarity between blocks can be quickly determined and the accuracy of the image similarity judgment can be enhanced. The experimental results show that compared with the grayscale histogram method,the proposed algorithm significantly improve the accurate-complete rating of image retrieval. Compared to the fractal retrieval algorithm in literature, the algorithm shortens the encoding time and reduces the number of sub-blocks, thereby improves retrieval efficiency.

关 键 词：分形编码 自适应四叉树 图像检索 相邻匹配 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]