Research And Design Of Real-time Image Dehazing System Based On FPGA

Affected by haze weather,the quality of images captured by outdoor imaging equipment is degraded,which greatly interferes with subsequent image processing.Therefore,image dehazing technology has received extensive attention.The dark channel dehazing algorithm has gradually become a research hotspot at home and abroad due to its remarkable dehazing effect.The dark channel dehazing algorithm processes images serially on the software platform,which has high algorithm complexity and low real-time performance.FPGA has the advantages of parallel high-speed processing and low power consumption,which provides a new research direction for real-time image dehazing.This paper deeply studies the single image dehazing algorithm based on dark channel prior theory.In view of the halo effect existing in the current algorithm,it is difficult to realize real-time image processing,a dark channel dehazing algorithm combining downsampling and Sobel edge detection is proposed.Channel dehazing algorithm.Firstly,a down-sampling module is added to reduce the computational load;Then,Sobel edge detection is introduced,and different filter sizes are used to divide different areas to obtain high-quality dark channel images,and use the row maximum mean value as the atmospheric light value to reduce deviations;Finally,the limited edge area is refined by guided filtering to reduce blockiness and restore the image.Based on the improved defogging algorithm and relying on the ISE development platform,a real-time defogging system framework is constructed.The OV5640 camera is configured to collect images through the IIC protocol,the SDRAM controller is designed to cache data,and the VGA driver controls the defogging image display.Combined with the improved algorithm,a general design scheme of defogging algorithm based on filter selection is proposed by modular design.Hardware logic,focusing on the analysis and discussion of the realization of the filter selection part.Finally,integrate the design of the FPGA-based hardware dehazing system,perform timing analysis and constraints on the system,test the down-board,read the hardware real-time dehazing data through the serial port,and perform images from three aspects:Information entropy,Peak Signal to Noise Ratio and Average gradient.Comparative experiments show that the improved dark channel dehazing system based on FPGA proposed in this paper can effectively improve the halo effect of dehazed images and restore more detailed information.At the same time,the system has real-time,greater flexibility and adaptability,and has certain application value in target tracking,remote sensing imaging and so on.