基于多光谱融合图像的背景分割

Background Segmentation Based on Multi-spectral Fusion Images

针对可见光或近红外图像背景分割存在精度不高的问题,提出多光谱融合的图像分割。首先采集红色、绿色和近红外3个波段光谱图像,采用r+ir、2g-r-ir、R+G+IR、g+ir、g+r、ir-r、IR-R等融合方式,然后使用最大方差自动取阈值法确定阈值,并根据此阈值分割图像。在试验中,进行了各种融合图像分割效果的对比研究,并与原始的单波段光谱图像进行比较研究。试验结果表明,IR-R的融合方式效果最好。

To overcome the low precision of background segementation problem in visible or IR image. Image segmentation based on multi - spectral images fusion is present in this paper. Firstly the R,G and IR images are acquired respectively , the r ＋ ir ,2g - r - ir,R ＋ G ＋ IR,g ＋ ir,g ＋ r,ir - r,and IR - R fusion methods are used , then the threshold is chosen based on the Otsu method, used to segment the image. Analysed the segmentation effect of different multi - spectral fusion methods and compared with the single image in experiment. Results show that IR - R fusion image get the best performance.