Research On Multiple Exposure Images Fusion

Nature presents us a colorful world due to high dynamic range, yet the image acquisition devices cannot fully grasp nature’s original appearance due to the limited dynamic range. The device has too short exposure time to capture all image details, some regions of image appears underexposed; In contrast, given more exposure time, image will appear overexposed that some regions are shown too bright. Thus, when the natural scenes with whole dynamic range is captured by the acquisition equipment with limited dynamic range, the multiple exposure image with diversification of details appears, included their individual image details. What if these multiple-detail images could be fused together, the final image will be a high quality result. How to apply a reasonable solution about fusion image is the focus of this thesis.On the basis of a large number of foreign language references, research on multiple exposure image fusion, this thesis introduces several image fusion algorithms in detail, the relevant simulations of pyramid image fusion and block image fusion are completed with Matlab. By comparisons, the suitable image fusion algorithm for hardware development model is adopted, the behavior simulation is finished, and the solution that can apply the multiple image fusion algorithms to hardware platform is proposed. Focus on data accuracy, comparison of design system is built, the instant result can be verified within the development process for the assurance of data correctness.The main points of this thesis are data storage system and algorithms processing system. With SDRAM chips as basic unit, data storage system provides solutions for the non-match clock between systems with different levels, some FPGA chip resources are released due to the optimized controller of data storage system, data storage system is very important to the whole design. Algorithms processing system is also focused on this thesis, the functional simulations are finished, all simulation results have been proved for fulfilling the basic algorithms requirements. Besides, algorithmic manipulation based on hardware pipelining processing method is proposed.According to the top-down design rule, modular design has been finished. With the help of FPGA IP resources, the processes of counting pixels, calculating the entropy values, weighting fusion units have been implemented by using Verilog hardware description language, ready for the final implementation of multiple exposure image fusion algorithms on FPGA platform.