郭二军
摘 要: 传统的图像匹配融合方法在匹配多重纹理图像时,很容易出现误差匹配,融合后的图像清晰度不高,轮廓不鲜明,针对上述问题,在云平台网络上研究了一种新的多重纹理图像匹配融合方法。首先,计算多重纹理图像的匹配代价,分析图像像素的相似度和特异性,构建动态规划路径,在不同网络结构下匹配多重纹理图像;然后,建立树状图对图像进行融合;最后,利用视察矫正方法将匹配融合得到的误差点消除。为验证该方法的工作效果,与传统匹配融合方法进行实验对比,结果表明,给出的方法能够清晰地得到像素点云,使融合后的图像轮廓鲜明,画质清晰,适用于图像重构。
关键词: 云平台; 网络图像; 多重纹理图像; 图像匹配; 图像融合; 融合方法
中图分类号: TN911.73?34; TP391.41 文献标识码: A 文章编号: 1004?373X(2019)19?0059?05
Abstract: The traditional image matching and fusion methods are prone to error matching when matching of multi?texture images is conducted. The fused image is not clear and its contour is not clear. To solve the above problems, a new multi?texture image matching and fusion method is studied on cloud platform network. Firstly, the matching cost of multi?texture image is calculated, the similarity and specificity of image pixels are analyzed, the dynamic programming path is constructed, and the multi?texture image is matched under the condition of different network structures. Then, the tree image is established for image fusion. Finally, the error points obtained by matching fusion are eliminated with inspection correction method. In order to verify the effectiveness of this method, some experiments are carried out for comparison with traditional matching and fusion methods. The results show that the proposed method can obtain the pixel point?cloud clearly, which makes the fused image contour distinct and quality clear, and is suitable for image reconstruction.
Keywords: cloud platform; network image; multi?texture image; image matching; image fusion; fusion method
图像纹理是能够表述图像表面和结构的基本属性,通过图像的平均亮度、最大亮度、最小亮度、图像尺寸、图形形状来描述[1]。纹理元素随机建立空间关系,经过一段时间,图像纹理之间的基元呈现相关性关系。多重纹理图像的匹配融合在创建逼真的三维模型中发挥着重要的作用,在广告、动画、视频等领域有着广阔的发展空间。目前研究的网络多重纹理图像匹配融合技术多是利用人机交互界面,虽然取得的图像精度很高,但是操作过程复杂,自动化效果差,在规划时仅能使用一条扫描线,很容易出现匹配错误,尤其是对于一些纹理不够充分或者是局部区域有重复特征的图像,目前方法的缺点更加明显[2]。
相较普通图像而言,多重纹理图像结构复杂,匹配融合更加困难。本文在云平台网络中分别对多重纹理图像的匹配方法和融合方法进行研究,内部设立了立体视觉系统,利用摄像机锁定目标,在三维网络和二维网络中完成匹配和融合工作,并将多重纹理图像的信息进行恢复。在平面视觉和立体视觉领域,图像匹配和融合是最关键的两个步骤[3]。利用匹配得到的视差图测量物体景深,在不同的约束条件下,有着不同的匹配和融合方法,分别是针对小区域进行匹配和融合以及针对全局进行匹配和融合[4]。多重纹理图像的小区域匹配融合工作要比全局匹配融合工作简单,产生的误差也小。本文引入图像重构算法,研究像素点与像素点之间的相似性,分析图像自身的特异性,使匹配的图像梯度不断加大,利用视差图矫正误差匹配点和误差融合点。
DENG Bo, MAO Hongyu, WANG Xiaofeng, et al. Emulation method for the fusion of virtual and physical networks supported by cloud platform [J]. Journal of Chinese computer systems, 2018, 39(3): 478?483.
[2] 沈蒙,程国华,祝烈煌.支持隐私保护的加密遥感图像融合算法[J].中国科学:信息科学,2017,47(2):736?751.
SHEN Meng, CHENG Guohua, ZHU Liehuang. Privacy?preserving encrypted image fusion algorithm for remote sensing [J]. Scientia sinica informationis, 2017, 47(2): 736?751.
[3] 石祥滨,钟刘倍,张德园.Hadoop环境下图像内容检索方法的研究[J].沈阳航空航天大学学报,2017,34(3):63?69.
SHI Xiangbin, ZHONG Liubei, ZHANG Deyuan. Research on the method of retrieval of image content based on Hadoop [J]. Journal of Shenyang Aerospace University, 2017, 34(3): 63?69.
[4] 赵妍,苏玉召.融合物联网云平台的智慧城市研究[J].教育教学论坛,2017(23):80?81.
ZHAO Yan, SU Yuzhao. Intelligent city research on the integration of internet of things cloud platform [J]. Education teaching forum, 2017(23): 80?81.
[5] 李冰.颜色纹理形状特征分层融合图像检索方法研究[J].西南师范大学学报(自然科学版),2017,42(1):54?59.
LI Bing. On Technology of computerimage retrieval based on multi feature fusion [J]. Journal of Southwest China Normal University (Natural science edition), 2017, 42(1): 54?59.
[6] 潘红艳.分层多特征融合图像检索方法研究[J].无线互联科技,2018,15(10):68?70.
PAN Hongyan. Research on stratified multi features fusion based on image retrieval method [J]. Wireless internet technology, 2018, 15(10): 68?70.
[7] 汪玉美,陈代梅,赵根保.基于目标提取与拉普拉斯变换的红外和可见光图像融合算法[J].激光与光电子学进展,2017,54(1):98?106.
WANG Yumei, CHEN Daimei, ZHAO Genbao. Image fusion algorithm of infrared and visible images based on target extraction and laplace transformation [J]. Laser & optoelectronics progress, 2017, 54(1): 98?106.
[8] 张晓闻,靳雁霞,银莉,等.融合粒子群与拓扑相似性的图像匹配算法研究[J].微电子学与计算机,2017,34(3):95?99.
ZHANG Xiaowen, JIN Yanxia, YIN Li, et al. Matching algorithm of the image based on particle swarm optimization and topological similarity [J]. Microelectronics & computer, 2017, 34(3): 95?99.
[9] 王旭东,王晓卫,李明哲.基于DM642嵌入式图像融合处理系统硬件设计[J].电子设计工程,2017,25(18):172?177.
WANG Xudong, WANG Xiaowei, LI Mingzhe. Hardware design of embedded video fusion processing system based on DM642 [J]. Electronic design engineering, 2017, 25(18): 172?177.
[10] 蔺素珍,韩泽.基于深度堆叠卷积神经网络的图像融合[J].计算机学报,2017,40(11):2506?2518.
LIN Suzhen, HAN Ze. Images fusion based on deep stack convolutional neural network [J]. Chinese journal of computer, 2017, 40(11): 2506?2518.
[11] 赵玲娜,花向红,刘闯,等.基于激光反射率的点云与图像自动融合[J].测绘工程,2017,26(8):29?34.
ZHAO Lingna, HUA Xianghong, LIU Chuang, et al. Auto?fusion of point clouds and image based on laser reflectance [J]. Engineering of surveying and mapping, 2017, 26(8): 29?34.
[12] 孙志田,张建梅,霍丽芳.基于最小生成树的图像融合算法[J].计算机仿真,2017,34(3):277?279.
SUN Zhitian, ZHANG Jianmei, HUO Lifang. Image fusion based on minimum spanning tree [J]. Computer simulation, 2017, 34(3): 277?279.
[13] 汪强,尹峰,刘钢钦.基于小波的彩色图像融合技术[J].计算机仿真,2017,34(11):201?204.
WANG Qiang, YIN Feng, LIU Gangqin. The Color image fusion algorithm using wavelet transform [J]. Computer simulation, 2017, 34(11): 201?204.