GOP Size Influence in High Resolution Video Streaming over Wireless Mesh Network Rafael Sousa, Edjair Mota, Edson N. Silva Jr., Kleverson Santana Paixão Computer Science Department Federal University of Amazonas (UFAM) Manaus, AM – Brazil {rlms,edjair,edson,ksp}@dcc.ufam.edu.br Bruno Faria Nokia Institute of Technology Manaus, AM – Brazil Bruno.faria @indt.org.br João Batista Pinto Neto Network Administration Department Public State Ministry of Law Porto Velho, RO - Brazil jbpinto@mp.ro.gov.br Abstract – Video transmission over IP networks requires appropriate conditions of quality of service, in order to ensure the integrity of the image, which still represents a challenge to be overcome in wireless mesh networks due to traffic capacity limitations inherent to its architecture. In this work we evaluate the performance of high-resolution video over multi- hop mesh networks and we propose a suitable GOP size for encoding video to be streamed, based on traffic conditions to minimize the loss of quality. Experiments were carried out in a wireless mesh network planned for real testing within a 1500m 2 area of the main UFAM campus, using the IEEE 802.11a and 802.11n network standards. Keywords - gop; video streaming; mesh I. INTRODUCTION Performance of transmission of real-time applications data has been the subject of research in telecommunications networks. Additionally, the importance of wireless mesh networks has grown in recent years due to its low cost of deployment and versatility. However, there are challenges associated with the delivery of multimedia services with quality and providing a good experience for the user in this type of network. In this article, we investigated the effect of different values of GOP (Group of Pictures), used in video compression, on the transmission quality of the video stream. Packet loss in transmission or high value of delay degrade the video quality and avoid the faithfully reconstruction at the receiver. Therefore, it is proposed a suitable value of GOP for transmission of video encoded in MPEG in a wireless mesh network composed of three jumps with channel sharing. The approach to change the GOP for a better video quality is studied in various forms. In [1] the main focus is on efficiency of video encoding and the approach aims to use this concept as a means to mitigate the loss of quality in transmission in a wireless network of multiple jumps. [2] presents a study on transmission of streaming video over multi-hop wireless mesh networks, evaluating the quality metric based on PSNR. The remainder of this paper is organized as follows. Section 2 describes some basic aspects of video encoding necessary for understanding the investigation. Section 3 describes the experiments scenario, as well as the main tools used. Section 4 explains the methodology of experimental design. Section 5 presents the results and analysis. Finally, section 6 presents the conclusion and proposed future work. II. VIDEO ENCODING The most used compression system for video currently employed was developed by the group MPEG (Moving Pictures Experts Group). It uses the concept of Discrete Cosine Transform (DCT) that calculates specific aspects for each frame of 8 × 8 pixels of a given intra-frame (or I- Frame) [3]. This kind of frame is more faithful to the original and does not require any time reference from any other frame. The two other kind of frames are the P and B (predictive and bi-directional, respectively), whose functions are to encode the temporal differences between consecutive frames of the original video based on I-frame. This technique takes advantage of the degree of similarity between consecutive frames. The P-frames are frames that are constructed from references to I and P frames. B- frames are those which have bidirectional reference, i.e., they are built from I or P frames immediately preceding and following them. When an I frame is encoded, we say that at that moment a Group of Pictures (GOP) begins, so a GOP ends when a new I-frame is found. From this concept, we can make the following assumption: if every frame I occupies more bits than other frames, then less I frames means smaller compressed video. Therefore, the higher the GOP, the smaller the size of the video in bits, and less disk space and bandwidth it occupies. 978-1-4244-7755-5/10/$26.00 ©2010 IEEE 787