基于ZYNQ的实时去雾系统设计实现

Design and implementation of real-time dehazing system based on ZYNQ

针对雾霾天气下视觉系统成像存在纹理色彩信息丢失、无法实时去雾等问题,提出一种基于可编程逻辑架构改进的Retinex去雾算法。在传统Retinex算法基础上进行光照估计、色彩平衡、图像融合等多方面优化,通过指令约束、逻辑优化满足系统资源要求。实验结果表明,相较基于HSV空间的多尺度Retinex算法处理,改进算法处理后3幅图像平均的信息熵和峰值信噪比略有提高,结构相似性指标至少提高27.01%。在图像处理耗时方面,有至少17倍的加速,处理一帧800×480大小彩色图像仅需4.03 ms。验证了构建的去雾系统同时满足了去雾效果与实时性的要求。

Aiming at the problems of the loss of texture and color information and the inability of real-time defogging in visual system imaging under haze weather,an improved Retinex defogging algorithm based on programmable logic architecture was proposed.Based on the traditional Retinex algorithm,illumination estimation,color balance,image fusion and other aspects were optimized,and the system resource requirements were met through instruction constraints and logical optimization.Experimental results show that,compared with the multi-scale Retinex algorithm based on the HSV space,the average information entropy and peak signal-to-noise ratio of the three images processed using the improved algorithm are slightly improved,and the structural similarity index is increased by at least 27.01%.In terms of the time consumption of image processing,it is at least 17 times faster,only takes 4.03 ms to process a frame of 800×480 color image.It is verified that the constructed defogging system meets the requirements of defogging effects and real-time performance at the same time.