透過您的圖書館登入
IP:3.140.185.170
  • 學位論文

結合動態攝影機的運動體偵測與嵌入式壓縮的編碼率與品質調整

Rate and Quality Control with Embedded Coding and Moving Object Detection of Mobile Cameras

指導教授 : 連豊力

摘要


近年來,行動式機器人因為具有主動與機動的特性,已被廣泛研究於各個領域。尤其在監測系統上的應用,機器人之間的合作使任務執行更有效率。因此,即時性影像資料的交換傳遞便成為重要的研究課題。然而,現今網路並沒有提供服務品質(quality of service)的保障,封包在異質網路(heterogeneity)上傳輸 ,常會因為阻塞而造成延遲與遺失,大大降低接收影像的品質。 影像傳送率的控制(rate control)可以避免網路阻塞的發生。過去的研究提出不同的架構設計來實行,藉由不同的編碼方式與傳輸方式在頻寬使用效能與彈性之間做權衡取捨。本文提出一種以接收端為基礎,品質保證的傳送率控制架構,主要是調整嵌入式編碼的編碼率與品質,期望利用群播(multicast)傳輸方式增進頻寬使用效能,也能使接收端依照收到的資料量而有不同程度的影像品質。 整體架構分為空間與時間取樣兩大方向。時間取樣利用邊緣多寡挑選出清晰的幀,可以減少相似與模糊幀的編碼與傳送的花費。在空間取樣上,本文提出一種可應用於動態攝影機的移動物偵測,藉由去除掉攝影機移動造成的背景變動,進而偵測出主動移動物體的範圍。接著,嵌入式編碼捨棄部分小波係數使影像具有多解析度的特性,並對主動移動物體的範圍優先編碼,以提供對此感興趣區域(region of interest)品質的保證。 本文模擬三種實驗情境,來驗證提出的架構中動態攝影機的移動物偵測的正確性,分析接收端在不同有限頻寬下影像的品質,並且統計頻寬的使用效能。

並列摘要


Because mobile robots are active and spontaneous, the studies on mobile robots have been more popular in recent years. Especially in the applications of surveillance system, the cooperation among security mobile robots can help finish task more efficiently. Thus, real-time video transmission and information communication become very important. However, there is no quality of service (QoS) guarantee for video transmission over the current Internet. Transmission over heterogeneous network usually causes congestion which leads to high packet loss ratio and delay and degrades the presentation quality of received video. Rate control has been proposed by past studies to prevent congestion occurs. Different types of rate control mechanisms coding and delivery in different methods to compromise the tradeoff between bandwidth efficiency and flexibility. The thesis proposes a receiver-based, rate and quality control. The main principle is to regulate encoding rate to guarantee the video quality is acceptable, especially in region of interest. Transmitting the encoded bit stream in multicast delivery not only improves bandwidth efficiency but also provides different quality for receivers with different length of bit stream. The overall mechanism can be separated into spatial and temporal sampling. Temporal sampling chooses the clearest frame to reduce the cost of encoding and transmitting blur or similar frame. On the spatial sampling, a method to detect moving object by eliminating the changing backgrounds caused by the motion of mobile camera is proposed. Then, embedded coding is performed to discard partial wavelet coefficients to provide multi-resolution of the frame. Encoding the region where there is moving objects in prior to the fixed background guarantees the quality in region of interest is acceptable. The thesis simulates in three scenarios. The objectives are to the correctness of moving object detection in mobile cameras, to analyze the quality of received video under different bandwidth limitation, and to examine the bandwidth efficiency with the proposed rate and quality control.

參考文獻


[1: Gonzalez and Woods 2008]
[3: Ghanbari 2001]
[4: Foresti et al. 2001]
G. L. Foresti, C. S. Regazzoni, and R. Visvanathan, “Scanning the issue/technology - Special issue on video communications, processing and understanding for third generation surveillance systems,” in Proceedings of IEEE, Vol. 89, No. 10, pp. 1355–1367, Oct. 2001.
[5: Luo et al. 2005]

延伸閱讀