Research On Image Defog Algorithm Based On System On-chip

Under haze conditions,degraded images captured in fog suffer from poor contrast,low visibility and color shift.This significantly affects the stability and accuracy of outdoor video systems.The traditional real-time image defogging algorithm is based on a PC.It's hardware cost and the power consumption is high.Therefore,researching on efficient,low-cost,lowpower and miniaturized real-time image defogging algorithm has great practical value.This paper proposes an improved Retinex image defogging algorithm,and implements the algorithm based on the system-on-a-chip.The main researches as follows:(1)The research and improvement of Retinex algorithmBased on the analysis of the basic principle of Retinex defogging algorithm,an improved algorithm of Retinex defogging is proposed for the shortcomings of the algorithm in real-time processing.The algorithm converts the image from the RGB(Red,Green,Blue)color space to the HSV(Hue,Saturation,Value)color space to reduce the computational complexity of the image defogging algorithm;the luminance component V performs the Retinex defogging process,and the saturation S is stretched to extend the dynamic range and maintain the hue H unchanged.It not only reduces the computational complexity of the algorithm,but also avoids the color distortion caused by the traditional Retinex algorithm.(2)Research and design of image defogging algorithm hardware platformSeveral commonly used image defogging hardware platforms are compared and analyzed,and an image defogging program based on the ZYNQ system-on-chip is proposed.The application of hardware/software co-design patterns has improved the development efficiency and shortened the development cycle;studied the acceleration mechanism of the ZYNQ image defogging algorithm and the idea of Field-Programmable Gate Array(FPGA)development based on the Intellectual Property(IP)core;passed the framework structure of the ZYNQ system-on-chip and the Advanced eXtensible Interface(AXI)bus.After in-depth study,the hardware platform implemented by the algorithm is given;and the register convention of communication between ARM and FPGA is formulated.(3)Engineering implementation of the Retinex defogging algorithm based on a systemon-chipThe hardware and software partition scheme and data format of the on-chip system for image defogging algorithm are given.All kinds of control tasks such as image defogging algorithm,image color space conversion,linear stretching,logarithm operation,and other simple arithmetic modules are performed by ARM;FPGA acts as a coprocessor to accelerate the algorithm and mainly complete the Gaussian kernel function and two-dimensional The image convolution is computed in parallel to estimate the ambient light illuminance and the FPGA part of the algorithm is encapsulated into an IP core.The AXI bus inside the system on chip is used to complete the data exchange between ARM and FPGA,which alleviates the communication bottleneck problem of traditional multi-core systems.The experimental results show that the method proposed in this paper achieves the effect of real-time processing on the premise of reducing the cost and power consumption,and it's processing speed achieves 26 frames/s at image size of 300×200.So the method proposed in this paper has the advantages of fast processing speed,small size,embeddability,portability,low power consumption,etc.,and can meet the performance requirements of outdoor video systems.