Research And Implementation Of Image Dehazing Algorithm Based On ZYNQ Platform

In heavy foggy weather,there are phenomena such as low clarity,decreased contrast,and color distortion in outdoor shooting images.These phenomena have a serious impact on subsequent processing.Therefore,research on image dehazing under haze weather is crucial.With the continuous development of emerging technologies,existing intelligent systems have put forward higher requirements for the clarity and real-time performance of image dehazing.Traditional image dehazing platforms can no longer meet the requirements.This paper presents a study on image dehazing algorithms and their implementation on Xilinx’s ZYNQ platform using the ALIENTEK development board featuring the xc7z020clg400-2 chip.The ARM+FPGA co-design approach is adopted to enhance the effect of image dehazing.The main research content of this paper is as follows:1.Optimize the dark channel prior algorithm.First,by simulating the image after the dark channel prior algorithm dehazing using MATLAB software,the problem of poor dehazing effect in the sky area of the image and halo phenomenon in the edge area of the image in the dark channel prior algorithm is identified.Secondly,based on the dark channel prior algorithm,the sub-block division and histogram equalization ideas of the CLAHE algorithm are fused to solve the atmospheric light value of the image;and the transmission rate is refined based on the central surround theory of the Retinex algorithm.The value of the transmission rate in the sky area is improved by adding an adaptive coefficient to the refined transmission rate.Finally,subjective and objective evaluations were conducted on the dehazing effect of the optimized dark channel prior algorithm.The results indicate that,compared to other dehazing algorithms,the optimization of the dark channel prior algorithm in this paper has significantly improved the dehazing effect in the sky area of the image.2.Implement the optimized dark channel prior algorithm on the programmable logic(PL)side of the ZYNQ platform.Use the hardware description language Verilog HDL for RTL description in Vivado software.And modularize the design during implementation,mainly divided into five modules: dark channel image,initial transmission rate,rough transmission rate,transmission rate,and image restoration.Use methods such as adding pipelines,converting floating-point numbers to fixed-point numbers,and replacing multiplication and division with shift operations to realize the optimized dark channel prior algorithm.Each module is functionally simulated.3.Build a dehazing system based on the ZYNQ platform.Four system function modules for image acquisition,image processing,image storage,and image display are designed on the ZYNQ platform.Based on the AXI4 bus and IP core call,a softwarehardware collaborative dehazing system is built.Through the simulation of the optimized dark channel prior algorithm and the testing results of the dehazing system development platform,it is proved that the optimized dark channel prior algorithm has a certain improvement in the dehazing effect,and can reduce the dehazing running time of a 640 × 480 resolution image from1.7 s on MATLAB to 25 ms on Vivado,greatly reducing the running time of the image dehazing algorithm.