Research Of Image Defogging Algorithm Based On Physical Model And FPGA Implementation

In the haze weather,the visibility of the atmospheric environment is reduced,and the video and images taken by outdoor imaging equipment will be seriously degraded,which will affect the recognition and information analysis of human eyes and machines.The image defogging is to remove haze from foggy image to improve the image contrast and make the image clear.With the rapid development of FPGA technology,the realization method of image def ogging is no longer limited to CPU and GPU.Through the deep research of image defogging method,this thesis proposes a fast image defogging algorithm based on physical model,and realizes it on FPGA hardware platform.In this thesis,the difficulties of i mage defogging technology are summarized,and the realization principles of atmospheric scattering physical model and current mainstream defogging methods are deeply studied,including Retinex theory,histogram equalization principle,dark channel prior th eory and contrast enhancement principle.On the basis of theory and experiment,this thesis proposes a fast image defogging algorithm based on physical model.The method of estimating global atmospheric light value and atmospheric transmittance in the algo rithm is introduced in detail.The algorithm in this thesis has the ability to identify the sky region to a certain extent,and can estimate the global atmospheric light value more accurately.The global atmospheric light value is added to the process of e stimating atmospheric transmittance,which can automatically adjust the image brightness and the degree of haze removal.Based on MATLAB,this thesis develops the algorithm prototype for the method of defogging,and combines the subjective evaluation metho d with the objective evaluation method to compare the defogging effects of different defogging methods.The experimental results show that the propo sed algorithm not only has a good defogging effect,but also has a low complexity,showing the characteristi cs of real time and universality in defogging.On the basis of the fast image defogging algorithm based on physical model proposed in this thesis,the circuit of defogging algorithm is developed and designed by combining the hardware structure characterist ics of FPGA.The circuit of defogging algorithm is composed of dark channel module,mean filter module,global atmospheric light value module,transmittance module,image recovery module and memory circuit.The circuit design and simulation verification an alysis of each function module are carried out in detail.Finally,the FPGA results and MATLAB results are compared to verify the correctness of the circuit design of this thesis.When defogging an image with a resolution of 800×600,it only takes 16.8ms a nd the power consumption of the circuit is 1.5W.By comparing with other defogging schemes,it can be seen from the performance indexes such as processing time,power consumption and resource consumption that the FPGA implementation scheme in this thesis i s suitable for low power consumption and real-time application scenarios.