【作者】 孙韶媛；

【导师】 张保民；

【作者基本信息】 南京理工大学 ， 光学工程， 2002， 博士

【摘要】 夜视技术作为军用探测技术中的关键在战争中占据着重要地位。对夜视图像准确地获取、快速高效地传输与增强融合处理一直是国内外夜视领域极为重视的一个研究课题。本文以夜视图像的特征为算法理论基础，以改善夜视图像质量与更高效地传输夜视图像为主要目的，提出了适用于夜视图像的增强、融合与压缩编码的理论模型与算法，并以实验摄取的微光与红外热像为对象进行具体实验与研究，取得了如下成果： 首先提出了考虑相邻像素间二维关系的二维直方图信息容量这一图像评价参数，用二维直方图信息容量理论分析了微光图像与红外热像的特征，为论文后面对微光图像与红外图像进行压缩编码及增强融合处理提供了算法理论基础。 提出了一种改进的基于小波嵌入零树编码的夜视图像压缩方法，针对夜视图像独有的特点，在兼顾压缩比与图像质量的原则下，得到了较好的微光图像与红外图像的压缩图像。 提出了基于二维直方图分析的二元子图夜视图像增强处理算法，对微光图像与红外图像进行了增强处理。该算法克服了通常的直方图均衡算法的难以控制图像整体增强效果的不足，得到了质量较好的增强后图像，提高了目标的识别率。 针对微光图像与红外图像的特点，提出了一种基于小波变换的夜视图像融合算法，在小波变换域用基于对比度概念的融合算法对微光图像与激光助视图像、微光图像与红外图像进行了融合处理。得到了优于源图像的融合后图像。 提出了一种基于小波变换的微光图像与红外图像的调制融合算法，用对比度调制与灰度调制方法分别对微光图像与红外图像进行了融合处理。 在彩色夜视的概念基础上，对微光图像与红外图像进行了基于RGB色空间的假彩色融合处理，针对本实验采集的双谱微光图像与热红外图像的实验数据，对基于像素级彩色融合处理TNO算法进行了改进，得到了更有利于人眼识别的融合后彩色图像。更多还原

【Abstract】 Night vision technology is promoted greatly by the needs from the military. The obtainment,transmission and enhancment of night vision image have become the most important researchareas of night vision technology field. In order to improve the night vision image quality andtransmit it effectively, in this dissertation, based on the characteristics of the night visionimage, several algorithms of night vision image compression, enhancement and fusion areproposed. Several research achievements are presented in this dissertation as follows:1. An image evaluating parameter-two dimensional histogram information capacity is proposed in this dissertation. Then the characteristics of low light level image and infrared image are analyzed in detail. This is the theory basis of the night vision image compression, enhancement and fusion algorithm.2. A night vision image compression algorithm based on improved embedded zerotree wavelet is advanced, both the compression ratio and compressed image quality are well considered.3. An algorithm for night vision image enhancement based on dualistic sub-image and two dimensional histogram analysis is brought forward in this dissertation. It can eliminate the drawback of traditional histogram equalization that the fine part is not easy to control and protect. This is meaningful for night vision technology.4. A night vision image fusion algorithm based on wavelet transform is put forward in this dissertation. Two kinds of original images-low light level image and laser assistant vision image, low light level image and infrared image are fused based on the multiresolution contrast of the source images in the wavelet transform domain.5. A modulating fusion algorithm based on wavelet transform is advanced. The low light level image and infrared image are fused using contrast modulating and gray level modulating method.6. A false color night vision image fusion algorithm based on RGB color space is proposed in this dissertation. Low light level bispectral images and infrared images are fused using improved TNO algorithm. It can indeed bring us a natural color image and the object can be recognized more effectively.更多还原