Low-light Image Enhancement Algorithm Research And Hardware Implementation

With the rapid development of high-definition surveillance and large-scale projects such as safe cities and smart cities,the quality of surveillance images has become one of the pain points of intelligent video surveillance systems.In an environment with low light or uneven light,the collected images are usually not clear enough.The colors of the image are dim and the image content cannot even be distinguished,which can't meet the viewing and practicality.For national security and other tasks required to complete the target detection and tracking or other functions are also not available.Therefore,the research on low-light image enhancement algorithm and its technical application has important value.The research content and results of this thesis are as follows:(1)Due to the uneven illumination in the low-illuminance environment,this thesis proposes the Retinex low-illuminance image enhancement algorithm based on the illumination map estimation to solve the problems of halo artifacts and insufficient extraction of the main structure of the image in the illuminance estimation.The algorithm first uses the L₂norm to process the maximum pixel value in the RGB channel to reduce the halo artifacts.Then improves the regular term of the RTV model to smooth the image texture details and accurately extract the main structure.Finally,through the improved adaptive Gamma correction factor performs brightness correction.The processing results of the proposed algorithm and several classic algorithms are subjectively compared and objectively evaluated,which proves the effectiveness of the proposed algorithm and has a large performance advantage.(2)Based on the illumination-reflection model emphasized by the Retinex theory,this thesis improves the traditional histogram equalization algorithm.First,convert the image from RGB to YCb Cr color space,and take its Y component as the initial light map.Secondly,the use of histogram equalization and Gaussian filtering to refine the illumination map not only avoids the phenomenon of excessive enhancement,but also avoids the problem that the coefficient matrix in complex algorithms is too large and to be easily implemented by hardware.Finally,through experimental comparison and analysis,the subjective and objective evaluation results of the image are better than the traditional histogram equalization algorithm.(3)This thesis uses Altera's Cyclone IV E series development board to build an FPGA-based image enhancement system.First,perform the hardware logic design of the sub-modules for the image enhancement algorithm.Then debug and simulate through Quartus II 13.0.Finally download it to the development board and display the processing results on the TFT screen to achieve low-light image enhancement.