Infrared And Low Light Level Image Fusion Of Real-time Processor Improvements And Algorithm

Image Fusion Technology, as an important part of the multi-sensor information fusion, isvery popular research area in recent years. Especially, fusion of infrared andlow-level-light image becomes more and more popular in the field of night vision. Theyreact on different wave bands of target radiation and reflection, so their informationoutputs have complementarity. By means of image fusion and utilizes the messagecomplementarity, we can spread the time cover and the space resolving power of thesystem.In this thesis, most of the research works focus on the amelioration of the imageprocessor and algorithms. We study the origin of infrared and low-level-light, contrast theircharacters. For remove noise from the image effectively, a improved median filteralgorithm based on weighted median is presented. Before performing image fusion, theimages to be merged must be registered. We propose a new algorithm based on the windowto register the images. Considering the limitation of the fusion methods using the wavelettransform to process a large quantity of image data, such as high computational complexity,need more memory and difficult to realize on-line fast image fusion processing, a newimage fusion based on weighted median pyramidal decompositions is presented tocombine multi-source images at pixel-level. Experimental results show that the proposednew fusion method provides a good performance and realizes to merge multi-sourceimages quickly. FPGA in the instrument of VHDL language and the chart of principle hasthe function of synchronizing video data from two channels, real-time fusion anddisplaying the results. On the base of widely investigation of the image processor researchstatus at home and abroad, a design scheme of system for the dual-channel real-time imagefusion is proposed. The kernel processor adopted is TMS320DM642, the latesthigh-performance fixed-point DSP.