摘要
Abstract
The article proposed a GPU-accelerated video processing scheme based on the strict requirements of real-time and stability of picture processing for broadcasting.By designing data block-based division strategy and asynchronous transmission mechanism,efficient video data stream processing was achieved;an improved bilateral filtering algorithm was used for adaptive noise reduction,and the processing efficiency was enhanced by GPU parallel optimization;a timestamp-based frame synchronization strategy and a multilevel latency control mechanism were designed to guarantee stable operation.The experimental results showed that the total delay of 1080p60 video processing was controlled within 12.1ms,the signal-to-noise ratio was improved by 3.6dB,the chromaticity phase error was reduced by 1.2°,and all the performance indexes remained stable in the 7-day continuous operation test,which met the requirements of broadcasting-level video processing.关键词
GPU并行计算/实时视频处理/自适应降噪/延迟控制/广播电视Key words
GPU parallel computing/real-time video processing/adaptive noise reduction/latency control/broadcast television分类
信息技术与安全科学
