基于FPGA的1080P低质视频实时增强系统

1080P Low Quality Real-time Video Enhancement System with FPGA

传统的图像增强算法对暗光、背光等低质图像的亮光区域有过度增强的不足,且算法复杂度较高,难以应用于嵌入式小型系统中。为此,设计并构建了一种基于Xilinx XC7A200T FPGA的1080P低质视频实时增强系统,在研究IMSR(Integrated Multi-Scale Retinex)算法及传统算法的过度增强等问题的解决途径基础上,提出了结合颜色估计方法的改进IMSR增强算法。该算法采用Xilinx XC7A200T FPGA来构建低质视频实时增强系统的处理框架,并优化改进了IMSR算法各个模块在Vivado HLS(Vivado High Level Synthesis)上的实现过程。为验证所建系统的有效性和可行性,搭建了实验环境并进行了应用测试。验证实验结果表明,所构建的增强系统实现了预期的1080P低质视频实时增强功能,基本解决了传统Retinex算法对1080P低质视频增强中的过度增强及"雾化"偏色问题,具有增强程度自适应调整功能,占用资源较少,增强效果良好。

Conventional image enhancement algorithms have an over-enhancement effect on low-light areas of low-light images such as dark light and backlight, which have high complexity and difficulties in applying in small embedded system. A 1080P low quality real- time video enhancement system with Xilinx XCTA200T FPGA has been designed and implemented, where the Integrated Multi- Scale Retinex （IMSR） and improved IMSR combined with color estimation model are contributed. Video processing architecture of the low quality real-time video enhancement system with Xilinx XCTA200T FPGA and the modules for improved enhancement algorithm have both been established in high level. The experiment system has been constructed to test and verify the effectiveness and feasibility of the established enhancement system. The experimental results show that the established system has solved the problems of excessive enhancement and color cast in traditional Retinex algorithm and enhanced degree of adaptive adjustment function, which takes up less resources and has good real-time performance and enhancement effect for 1080P low quality video.