基于四波段图像融合的彩色夜视方法研究

为了在低照度条件下获取目标的彩色图像,提高夜间对目标的识别能力,提出了基于四波段图像融合的彩色夜视方法.采用F-P滤光片设计出了透射中心在三基色光中心波长及近红外波段的四波段滤光片,在各个透射区域的平均透射率均达90%以上;将四波段滤光片设计成圆形滤色轮结构,用分光计测得了各波段滤光片的光谱透射性;对加入滤光片后的系统信噪比进行了分析和计算,分析结果表明加入滤光片不会引入噪声,计算结果表明加入蓝色、绿色、红色和近红外波段滤光片的成像系统的信噪比分别是原来单色微光夜视系统信噪比的19.59%、38.45%、47.28%和46.70%.借助国产超二代像增强器在微光实验室进行了四波段图像采集及彩色图像融合实验,实验时光照度分别为1×10~(-3)lx和1×10~(-1)lx,对获取的图像质量进行了评价.结果表明:在照度为1×10~(-3)lx时,融合的彩色图像在均值、方差和熵这三项指标上均优于过滤后的蓝色和绿色的单色图像,且由于彩色图像中利用了近红外图像进行增强,使得彩色图像亮度更高,颜色分辨性更好;在照度为1×10~(-1)lx时,融合后的彩色图像的信息熵比红、绿、蓝三种基色图像的大,彩色图像携带的信息量更大.本文的研究对彩色夜视成像系统的设计和研发具有借鉴和指导意义.

Study on Color Night Vision Image Fusion Method Based on Quadruple-band Images

In order to acquire color images of targets under Low Light Level(LLL) and improve the ability to distinguish targets from background at night,a color night vision image fusion method based on quadruple-band images was proposed.The quadruple-band filter with a transmittance at the center wavelength of three-band color lights and near-infrared band was designed by using Fabry-Perot(FP)filter,and the average transmittance is up to 90 ％ at each transmission region.The quadruple-band filter was designed as a color-wheel filter structure,the spectral transmittance of each single band filter was measured by a spectrometer.The Signal Noise Ratio(SNR) of the night vision system after adding the filter was analyzed and calculated.The analysis results show that adding the filter will not introduce addition noise,and the SNR of the imaging system which added the blue,green,red and near-infrared band filters is equivalent to 19.59％,38.45％,47.28％ and 46.70％ to compare with that of the original monochromatic night vision system,respectively.Using a domestic Super Gen Ⅱ image intensifier,quadruple-band image acquisition and color image fusion experiments were carried out in LLL laboratory,the experimental illumination is 1 × 10-1lx and 1 × 10-3lx,a set of images was adopted to conduct quality assessment.Experiment results show that,when the illumination is 1 × 10-3lx,the fusion color image is superior to the filtered blue and green monochromatic images both in gray mean,gray variance and information entropy,and the color image is enhanced by the use of the near infrared image,as a result the color image brightness is higher and the color identification is better.When the illumination is 1 × 10-1lx,the information entropy of the fusion color image is bigger than that of the red,green and blue color images.This research can give reference to the design and development of color night vision system.