Woolven S, Ristic V M, Chevrette P
Appl Opt. 1993 Nov 10;32(32):6556-61. doi: 10.1364/AO.32.006556.
A unique hybrid optical-digital image-processing system that functions at real-time rates and performs analysis in Radon space is presented. This system functions by using the forward Radon transform (a mathematical tomographic transform of image data from two-dimensional image space to onedimensional Radon space), which is achieved by a front-end optical processor followed by a digital processing subsystem operating in Radon space. The system works by optically converting the two-dimensional image data into a series of one-dimensional projections. All further processing is performed digitally in Radon space on the one-dimensional projections. Using the system in transform space, we show that it can perform real-time detection of minimum-resolvable-temperature-difference measurement targets better than a human observer. Also, this paper discusses the potential of real-time object-moment analysis in Radon space. These object moments can be calculated in Radon space with significantly less image data and fewer digital processing operations than in image space. The optical front end is capable of performing 6.04 × 10(10) operations/s on the two-dimensional image data.
本文介绍了一种独特的混合光学 - 数字图像处理系统，该系统能够以实时速率运行，并在拉东空间中进行分析。该系统通过使用正向拉东变换（一种将图像数据从二维图像空间转换到一维拉东空间的数学断层变换）来运行，这是由一个前端光学处理器实现的，随后是一个在拉东空间中运行的数字处理子系统。该系统通过将二维图像数据光学转换为一系列一维投影来工作。所有进一步的处理都在拉东空间中对一维投影进行数字处理。在变换空间中使用该系统，我们表明它能够比人类观察者更好地实时检测最小可分辨温差测量目标。此外，本文还讨论了在拉东空间中进行实时物体矩分析的潜力。与在图像空间中相比，这些物体矩可以在拉东空间中用显著更少的图像数据和更少的数字处理操作来计算。光学前端能够对二维图像数据执行6.04×10(10)次操作/秒。
