Key Technologies Of Infrared And Low-Level-light Image Real-time Fusion

In the modern war, whether finding targets more quickly, earlier and more precisely is the most important factor deciding the outcome of the war. By sensors' cooperative work, multiple spectral fusion techniques can capture more information; obtain various characteristic of the goal scene to the ultimost extent. In the hidden and disguised battle scene, low-light-level (LLL) and infrared fusion system can detect targets more quickly and more accurately, therefore, it is of major practical significance to develop real-time infrared and LLLimage fusion system for the progress of our national defense. In this paper, infrared and LLL image fusion theory, the design of optical and mechanical structure of the infront detecting system, real-time signal process of fuion imaging have been carried out, centering on realtime infrared and LLL image fusion.Image preprocessing is the precondition of image fusion. The imaging principle and image feature of LLL image and infrared image have been analyzed. The theories of recursive denoising and wavelet transform are adopted to denoise the LLL images and the algorithm of adaptive time-domain recursive denoising is put forward, which has remarkable effects on LLL image denoising and better realtime performance. The statistical principle of image gray histogram is applied to enhance the contrast of infrared images and the algorithm of adaptive piecewise linear transformation is come up with, which not only can reinforce the layers of infrared images but also has good realtime performance and is easy to realize on hardwares. Based on studying image fusion algorithm and image registration, a window fusion technique has been introduced, which can improve the realtime performance of the fusion system and solve the contradiction between image fusion quality and fusion speed. The feasibility of window fusion technique is validated using 2-D fast Fourier transform (2-D FFT) in frequency-domain. As a special case of window fusion, multi-source image fusion in different fields-of-view is discussed in detail in this paper.False color image fusion in YCbCr color space has been studied as well. Applying the principles of color transform and color transfer in color space, color transfer fusion algorithm in YCbCr color space, adaptive false color fusion algorithm and reginal coloring false color algorithm are introduced, which can well transfer the environmental image color to gray fusion image. Adaptive false color fusion algorithm adopts the data look-up table method to reduce arithmetic labor, which saves the system uptime and ensures the system's realtime performance. The false color fusion in the fusion system mentioned in this paper adopts this algorithm.By studying the above-mentioned theories, this paper has developed LLL and infrared realtime fusion system. The precisely designed double-channel parallel optical axis optical system with optical axis alignment device ensures the imaging quality; the FPGA-centered image preprocessing circuit can realize such application functions as video capture and image enhancement; the DSP-centered fusion circuit includes basic function modules like video synthesis, keyboard control and serial communications. The real-time image preprocessing program designed with VHDL and running on FPGA improves the quality of source images; the application programs designed with C and running on DSP such as image histogram statistics, gray fusion and false color fusion realize the real-time fusion of double-channel night-vision images. The real time operation system DSP/BIOS is adopted to optimize the design of softwares and to dispatch system resources reseaonably, which enhances the reliability of the system and improves the real-time performance of the system.