您的瀏覽器不支援JavaScript語法,但是並不影響您獲取本網站的內容。

技術探索

應用於即時串流之GPU硬體編碼器分享技術

中文摘要

自己的生活。使用者拍攝的視訊,經網路上傳到網路主機分享給許多觀賞者,而在觀看的同時,也可以藉由在視訊內容中加入文字、加入商標/浮水印和解析度轉換等功能,達到視訊內容的加值。但是這樣視訊編碼需要大量的運算資源才能即時處理多路直播串流,這會造成其運算成本大幅增加。因此,顯示卡廠商NVIDIA在顯示卡中放置專門為視訊編碼所使用的硬體編碼器,它可以提供比CPU更強大的編碼能力。但是,低價NVIDIA顯示卡限制只能同時編碼二路視訊,無法有效利用其編碼能力。本篇論文提出一顯示卡硬體編碼器分享方法與一串流排程方法,可有效地利用硬體資源達到多路即時視訊串流編碼。實驗結果顯示我們提出的方法可達到14路HD(High Definition)視訊即時編碼。

Abstract

Nowadays, based on the fast development of network and video codec technology, we can easily share our life through live video streaming. The live video streams generated from cameras are uploaded to remote video servers and shared to many on-line audiences. To increase the value of video content, the video servers can add texts, trademarks, watermarks, enhance video quality, and resize video resolutions. The video content after content processing needs re-encoding and the computation requirement is large. Therefore the cost that the video servers can encode multiple video streams in real-time increases significantly. Hence, NVIDIA, a graphics processing unit (GPU) manufacturer, has designed a dedicated hardware in its GPUs for video encoding. This encoding hardware can provide high video encoding performance compared to the conventional CPU encoding. Nonetheless, the number of concurrent video streams encoded by a low-end GPU is limited to 2, which cannot fully utilize its encoding power. In this paper, we propose a hardware sharing method and a scheduling method to efficiently utilize the encoding resource for encoding multiple real-time streams. The experimental results show that the proposed methods can achieve concurrent 14 HD (High Definition) video encoding in real-time.

關鍵詞(Key Words)

即時視訊編碼 (Real-Time Video Coding)
圖像處理單元 (GPU)
硬體分享 (Hardware Sharing)

相關檔案

共有0則留言張貼留言
顯示更多回答
歡迎留下您的意見:
姓名:
E-Mail:
文章留言:
輸入驗證碼:
請輸入驗證碼FK50