Research On Real-time Dehazing Algorithms For Videos And Its FPGA Implementations

Images collected by outdoor vision equipments usually have bad contrast and less details when it is in hazing weather conditions.It makes good importance to clear the hazing images and videos for the image information and the target in it.Up to now,most dehazing processings are performed on PC platform.To meet the requirement of real-time processing,a real-time dehazing system was designed based on FPGA platform in the thesis.At first,single image haze removal algorithm was studyed using dark channel prior.Due to the poor sky processing of the algorithm,an adaptive tolerance method was proposed to avoid color distortion in the sky region.Besides,the thesis achieved the refinement of the atmospheric transmissivity rapidly by using guided image filtering.For realizing real-time videos dehazing well,the calculating processings of the environment value and the transmissivity were optimized.Then the workflow of real-time video dehazing algorithm based on PC was provived.Secondly,based on the above algorithms,a real-time video dehazing system was designed and built based on FPGA platform,including a visible light CCD camera for shooting haze videos,a TVP5150 video decoder chip,a DDR3 SDRAM for caching videos,data,a Xilinx FPGA chip for dehazing algorithm processing,a TFP410 video coding chip and a HDMI HD display.As the core processor,the FPGA chip finished the Verilog design of dark channel prior,transmissivity estimation,guided filtering,and environment value.At last,haze images,videos and outdoor scenes were tested using the real-time video dehazing system.The experimental results showed that the system can achieve good effect on the images and videos dehazing.And it took only 0.75ms to dehaze a frame with 720*576 resolutions,so the system met the real-time requirement wonderfully.