多区域图像重建算法.pdf

Jo啪al puter Applications 计算机应用,2013,33(12):3544—3547,3566 ISSN lool一908l CODEN JYⅡDU 2013— 号:1001—9081(2013)—04 doi:.1001- 多区域图像重建算法万金梁1。,叶龙2 (,郑州450002; ,北京l00024) (+通信作者电子邮箱waIlqq-2006@) 摘要:为在多区域轮廓曲线中实现中等水平视觉的区域纹理替换,提出了一种多区域图像重建算法。该算法首先提取原始图像的分割区域及其索引号列表;其次根据分段迭代曲线拟合方法获得代表区域结构特征的多项式系数,并选取最小的区域纹理样本;最后利用重建的区域轮廓与合成的区域纹理实现区域恢复。该算法能成功实现包含大量纹理的自然场景图像重建,实验结果表明降低了图像重建所需数据量,适用于图像与视频的静止背景重建。关键词:人类视觉系统;图像纹理替换;纹理合成;曲线拟合;多区域中图分类号:’ 文献标志码:A Mlllti-region illIage reconstruction algoriuun WAN Jinliang”,YE hng 2 ( o,co唧l‘衍口以删钿Mt如nE嚼n8eri%胁Mn‰沁”渺o,&D加,,如n以k埘,肌蚴础胁Mn 450002,c^im; ^oDf矿蜘删如n西倒,聊一,lg,cDmmun泐如n‰泐瑙渺o,劬i∞,&班昭100024,miM) Akitract:Tb achieve medium levelvisualre舀on texture substimtion inmulti·re舀ons contour cunres, amulti—re西on image reconstmction algoritllm was pmposed. Firstly, thesegmented re西on锄d index ofori百nal image were ex讹cted. Secondly,the polyllomial coemcients representing re舀on stlllcture feature were got by piecewise ite枷ve curve fitting andthe sm0Uest re百on texture s啪ple was ,the recons协Jcted re舀on corltour锄dsylltllesizedre西on texture were used achieve re矛onal proposed algorithm can be done essfmly forimages witllmassive textures. The experimental resultsshow that“can reducethe锄ount ofdatarequired byimage reconstmction and beapplied toreconstmct th