提升图像采集系统的性能；针对不同视场的多-必读.docVIP

摘 要 现实生活中存在大量不同特性的不同视场的图像系统，不同视场的多源图像融合技术能够在不增加硬件设备投入的情况下，提升图像采集系统的性能；针对不同视场的多源图像融合技术，本文提出一种仿射变换和线性插值相结合的配准方法；对融合前后的图像在时域、空域进行了分析比较，并对融合的结果进行了评价和对比，验证了此技术应用的可行性。 Summary：There are a lot of different real-life characteristics of different imaging system field of view, different field of multi-source image fusion technology can not increase in case of hardware inputs, to enhance the performance image acquisition system; for different field of multi-source image fusion technology, this paper presents a affine transformation and linear interpolation method of combining registration; on the image before and after integration in time domain, spatial domain analyzed and compared, and the results of integration were evaluated and compared to verify this technical feasibility of the application. 关键词 多源图像 不同 视场 配准 融合 Keyword multi-source image, different, field of view, registration, fusion 引言 Introduction 像素级[1]融合是直接对源图像的像素数据进行融合处理，能更多地保留场景的原始信息，是应用比较广泛的图像融合方法。对源图像配准、视场的重合度要求较高，一般来讲在源图像的光学系统视场相同的情况下，采集到的图像更容易实现配准，也更有利于融合图像质量提高，也保证了融合图像的信息量。 Pixel level [a] fusion is directly on the source image pixel data fusion can be more reserved scenes of the original information, is used widely in image fusion. The source image registration, as the market require a higher degree of coincidence, in general, the optical system in the source image under the same field of view, images are easier to implement registration, and more conducive to integration of image quality, also ensure the entropy of the image. 但在实际应用中，往往会有一些多源图像采集设备的光学系统视场相差比较大，而每个源图像有各有各的特点（比如：可见光图像、微光图像和红外图像），融合图像更能体现场景的细节、突出场景目标。对于这种情况，在图像配准算法和图像融合算法的应用上又有一些不同特点。图1是不同视场图像融合示意图，两套不同视场的光学系统采集到得图像分别是图像A和图像B，图像B只对应于图像A中一部分，图像C是图像A和图像B的融合结果。实际上图像C中只有虚线部分才是融合图像，而其余部分只有图像A的信息。 However, in practice, often there will be some multi-source image acquisition device optical system field of view relatively large difference, and each source image has its own different characteristics (such as: visible light images, low light level i