【作者】 王辉；

【导师】 马泳；

【作者基本信息】 华中科技大学 ， 电路与系统， 2017， 硕士

【摘要】 图像融合技术可以融合不同信源获取的图像,使融合图像能够克服单一信源图像的局限性,保留各自的成像优势,提供更多有用信息,因此具有极大应用价值。红外图像与可见光图像由于成像机理差异,图像性质迥异,两者融合难度虽大,但优势明显。红外与可见光图像融合已在军事、安防、勘探、监控等领域得到应用,关于这项技术的研究也得到了更多关注。然而红外与可见光图像融合技术在理论和应用方面依然面临着亟待解决的问题。图像融合理论发展至今已取得突出成果,但是针对红外与可见光图像融合的算法却很少。红外与可见光图像融合技术的应用也因为实时处理能力的不足而受到制约,融合算法的计算复杂度高,融合设备计算能力难以满足实时处理要求,使得很多效果优秀的融合算法得不到应用。CUDA的出现使数据并行计算速度大幅提高,基于CUDA实现红外与可见光图像融合可以大大提升实时处理性能。本文在分析研究现有融合算法基础上,综合比较融合算法的计算复杂度,以及应用于红外与可见光图像融合时的融合效果,选取理想的算法并提出改进方案,同时设计了一套实际应用时的融合流程。融合算法的实现以GPU为计算核心,使用CUDA运算平台实现实时融合。在研究GPU的硬件架构及CUDA编程模型后,对算法细致划分计算任务,以CUDA逐一实现并优化加速,分析加速性能。最终,在实际场景中测试系统性能,实现了效果优秀的红外与可见光图像实时融合。更多还原

【Abstract】 Image fusion technology can fuse the images obtained by different sources,so that the fusion image can overcome the limitations of single source image,retain the respective imaging advantages,provide more useful information,so it has great application value.Infrared images and visible images due to differences in imaging mechanism,the image of different nature,the two integration is difficult,but the obvious advantages.Infrared and visible light image fusion has been in the military,security,exploration,monitoring and other fields have been applied,the study of this technology has also been more attention.However,infrared and visible image fusion technology in the theory and application is still facing urgent problems to be solved.Image fusion theory has made outstanding achievements so far,but for infrared and visible image fusion algorithm is very small.The application of infrared and visible image fusion technology is also limited by the lack of real-time processing capability.The computational complexity of the fusion algorithm is difficult to meet the requirements of real-time processing,so that many excellent fusion algorithms are not applied.CUDA makes the data parallel computing speed greatly improved,based on CUDA infrared and visible light image fusion can greatly enhance the real-time processing performance.In this paper,based on the analysis of existing fusion algorithms,the computational complexity of the composite fusion algorithm and the fusion effect of infrared and visible image fusion,we selected an ideal algorithm and proposed improvement method,while the design of a practical application of the integration process.The realization of fusion algorithm to GPU as the core computing,the use of CUDA computing platform for real-time integration.After studying the hardware architecture of the GPU and the CUDA programming model,the algorithm is carefully divided into the calculation task,and the CUDA is realized and optimized to accelerate and analyze the acceleration performance.Finally,in the actual scene test system performance,to achieve the effect of excellent infrared and visible image real-time fusion.更多还原