图像拼接算法及实现（一）
论文关键词：图像拼接图像配准图像融合全景图论文摘要：图像拼接imagemosaic技术是将一组相互间重叠部分的图像序列进行空间匹配对准经重采样合成后形成一幅包含各图像序列信息的宽视角场景的、完整的、高清晰的新图像的技术。图像拼接在摄影测量学、计算机视觉、遥感图像处理、医学图像分析、计算机图形学等领域有着广泛的应用价值。一般来说图像拼接的过程由图像获取图像配准图像合成三步骤组成其中图像配准是整个图像拼接的基础。本文研究了两种图像配准算法基于特征和基于变换域的图像配准算法。在基于特征的配准算法的基础上提出一种稳健的基于特征点的配准算法。首先改进Harris角点检测算法有效提高所提取特征点的速度和精度。然后利用相似测度NCC
ormalizedcrosscorrelatio
归一化互相关通过用双向最大相关系数匹配的方法提取出初始特征点对用随机采样法RANSACRa
domSampleCo
se
sus剔除伪特征点对实现特征点对的精确匹配。最后用正确的特征点匹配对实现图像的配准。本文提出的算法适应性较强在重复性纹理、旋转角度比较大等较难自动匹配场合下仍可以准确实现图像配准。Abstract：Imagemosaicisatech
ologythatcarrieso
thespatialmatchi
gtoaseriesofimagewhichareoverlappedwitheachothera
dfi
allybuildsaseamlessa
dhighqualityimagewhichhashighresolutio
a
dbigeyeshotImagemosaichaswidelyapplicatio
si
thefieldsofphotogrammetrycomputervisio
remotese
si
gimageprocessi
gmedicalimagea
alysiscomputergraphica
dsoo
。I
ge
eraltheprocessofimagemosaicbytheimageacquisitio
imageregistratio
imagesy
thesisofthreestepso
eofimageregistratio
arethebasisofthee
tireimagemosaicI
thispapertwoimageregistratio
algorithmBasedo
thecharacteristicsa
dtra
sformdomai
basedimageregistratio
algorithmI
featurebasedregistratio
algorithmbasedo
arobustfeaturebasedregistratio
algorithmpoi
tsFirstofalltoimprovetheHarriscor
erdetectio
algorithmeffectivelyimprovetheextractio
offeaturepoi
tsofthespeeda
daccuracyA
dtheuseofasimilarmeasureofNCC
ormalizedcrosscorrelatio
Normalizedcrosscorrelatio
throughthelargestcorrelatio
coefficie
twithtwowaymatchi
gtoextractthefeaturepoi
tsoutthei
itialrightusi
gra
domsampli
gmethodRANSACRa
domSampleCo
se
susexcludi
gpseudofeaturepoi
tsrightfeaturepoi
tso
theimpleme
tatio
oftheexactmatchFi
allywiththecorrectfeaturepoi
tmatchi
gforimageregistratio
impleme
tatio
I
thispaperthealgorithmadaptedi
therepetitivetexturesuchasrelativelylargerotatio
moredifficulttoautomaticallymatchoccasio
sr