CAJ | 학술논문

基于特征的图像匹配算法具有更广的适用范围和更好的性能表现，但因为其算法更加复杂，使其很难满足实时性，尤其当图像尺寸变大后，这一缺点更加限制了此类算法在实际工程中的应用。针对这一问题，本文提出了一种新的用于大幅面图像的快速匹配算法。算法采用空域分割和频域压缩的方式对图像进行预处理。通过提取目标图中高频信息丰富的区域作为待匹配子图，减小用于匹配的目标图像尺寸；通过小波变换提取源图尺度空间的高尺度表示，压缩 SIFT 特征点在图像频域中的存在空间。算法采用由粗到细的匹配策略，粗定位待匹配子图在源图中的空间区域后，再次进行细匹配操作，最终实现大幅面图像间的快速匹配。仿真实验表明，新提出的算法极大地提高了大图像匹配的速度，在对部分算法模块进行硬件加速后，新提出的算法甚至可以满足实时性的要求。
기우특정적도상필배산법구유경엄적괄용범위화경호적성능표현，단인위기산법경가복잡，사기흔난만족실시성，우기당도상척촌변대후，저일결점경가한제료차류산법재실제공정중적응용。침대저일문제，본문제출료일충신적용우대폭면도상적쾌속필배산법。산법채용공역분할화빈역압축적방식대도상진행예처리。통과제취목표도중고빈신식봉부적구역작위대필배자도，감소용우필배적목표도상척촌；통과소파변환제취원도척도공간적고척도표시，압축 SIFT 특정점재도상빈역중적존재공간。산법채용유조도세적필배책략，조정위대필배자도재원도중적공간구역후，재차진행세필배조작，최종실현대폭면도상간적쾌속필배。방진실험표명，신제출적산법겁대지제고료대도상필배적속도，재대부분산법모괴진행경건가속후，신제출적산법심지가이만족실시성적요구。
The matching algorithms based on features have more extensive adaptive range and better performance,because of more complexity,the ones are hard to meet real time,especially for big images,the disadvantage of non-real time further limits the application of this kind of algorithms in engineering practice.Aiming at this problem,a new fast matching algo-rithm for big image is proposed.The ways of spatial-domain segmentation and frequency-domain compression are applied to process images primarily.A region that contains abundant high-frequency information is extracted as sub-image of matc-hing,then the size of target image is cut down.A high scale image of source image in scale space is extracted via wavelet transform,then the range of SIFT feature points existing in frequency domain is compressed.Algorithm adopts a coarse to fine matching process,after coarsely locating the position of sub-image,another operation of refined-match is applied,fi-nally the fast matching for big image is completed.Simulation results testify that proposed algorithm cut down matching time greatly,after some modules are implemented on hardware system,the algorithm even meet the requirement of real time.