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

Due to the scattering effect of moisture on light in foggy days,the image acquired by outdoor image acquisition system has the disadvantages of low contrast,color distortion and detail loss.In order to reduce the impact of fog on the system and stabilize the acquisition of clear images,this thesis studies and improves the key algorithms in the current image defogging field,and designs a set of real-time image defogging system with wide application range and good defogging effect based on the speed advantage of FPGA digital circuit parallel computing.First,the basic architecture of the defogging system was designed.The IIC protocol was used to configure CMOS camera to obtain the original foggy image with a resolution of 1280×720 and 60Hz.The SDRAM control circuit was designed to cache the defogging image data,and the VGA control circuit was designed to display the defogging image.Secondly,to solve the problem of applying different defogging algorithms to light fog and heavy fog,it is proposed to use the size of image variance value to distinguish fog scenes,so that the defogging system has better adaptability.Then,two kinds of defogging algorithms,local histogram equalization for light fog scenes and dark channel prior for heavy fog scenes are studied.Aiming at the problem of easy distortion of composite image after RGB single-channel equalization of local histogram equalization algorithm,the method of conversion to YCBCR color space is adopted to equalize only the brightness components,thus solving the problem of easy distortion of defogging image.In view of the high complexity of the dark channel prior defogging algorithm and the difficulty in FPGA implementation,the fast minimum filtering was used to obtain the dark channel image,the threshold limiting method was used to quickly eliminate the sky region,the atmospheric light was obtained,and the second Sobel edge detection was used to limit the guided filtering range and improve the guided filtering speed.Finally,image grayscale and image variance circuits are designed on FPGA chip to distinguish fog scenes.Four circuits including histogram statistics,accumulation,equalization and zeroing are designed to complete the local histogram equalization algorithm.The dark channel prior algorithm is realized by the designation of four circuits including dark channel extraction,edge detection,custom division and image restoration.Sequence analysis and constraint were performed on the FPGA system,and real-time image defogging test was conducted in two fog scenes,heavy fog and light fog in the outdoors.Several frames of image data of hardware real-time defogging were transmitted back to PC through serial port,and image quality evaluation was conducted through information entropy,average gradient and variance.By comparing and analyzing the quality of the hardware real-time image after fog removal with the original image,the results show that no matter the external scene is heavy fog or light fog,the image after fog can be effectively recovered in real time.For 1280x720 resolution images,variance size can be increased by more than 10%after defogging.The whole time of the algorithm for fog removal has reached the nanosecond level.The experimental results show that the designed real-time image defogging system has a wide range of application,good defogging effect and certain practical value.