AN EFFICIENT ADAPTATION OF UHD VIDEO STREAMING OVER INTERNET

The main challenge of multimedia applications is how to transmission the Ultra-High Definition (UHD) video streaming in real-time over the internet. Real-time video streaming suffers from difficulties to be the flexible and efficient cause of the wide variation of the available internet bandwidth. To avoid the problems that introduce with the internet, in this work the HEVC with required video network adaptive streaming are proposed and tested using different six levels of three UHD video (4K, FHD, 720p, 4CIF, CIF and QCIF). Different experiments are applied to find the optimal configuration of H.265 encoding features for six levels to obtain the required PSNR with a range of 32-38 dB. The important part of this paper is a controller that worked incorporate with the encoder (H.265) to obtain the video streaming adaptation on the available bandwidth of the channel. The controller continuously reads the status of the channel buffer, then choosing the proper level of video to be transmitted over the channel. The work architecture is content two parts: First, the H.265 codec that applies on the three raw videos with optimal parameters configuration to compress them and get videos with lower bit rate and acceptable quality. Second, the compressed videos, based on controller selection, will be transmitted over two types of networks: Unicast and Multicast, and monitoring these two networks to evaluate their performance by obtaining different maps to choose the good one based on their effects on all levels of videos that are tested in this work. keywords: UHD video, HEVC/H.265, RTP protocols, QoS, Unicast, Multicast.


Iraqi Journal of Information and Communications Technology(IJICT)
Conference Series: The 1 st Conference of Applied Researches in Information Engineering(ARIE2021), 2021 ISSN:2222-758X e-ISSN:  prediction values to obtain restored pixel values, and these values are used in intra-prediction with a current frame of video [11]. The results from these operations pass through the post-processing. Then the restored value and post-processed frame are saved into Decoded Picture Buffer (DPB) that is used for inter prediction of more frames [12]. The mechanism of video measurements are objective and subjective techniques, what is applied in this work is the objective one. The calculation of objective measure was represented by measuring the Peak Signal-to-Noise Ratio (PSNR) and Bit Rate (BR) [13]. PSNR is used to calculate the quality of video sequence as an objective measure and it represents the average value of PSNR at each frame, it's found by the following equation [14]: Where the MAX represents the maximum value of the pixel, it's equal to (2n − 1) and the value of n is the bit number, while the MSE was the Mean Square Error found by the following equation [14]: Where M and N represent the picture size with width and height respectively, while the values of P anchor and P test are a single-pixel (the first is of the raw picture and the second is of compressed picture) [14].

II. THE PROPOSED SYSTEM OF ENCODING
The proposed system in this work mainly includes the controller that is incorporated with H.265 codec. This system presents reservation of bandwidth for UHD video streaming especially for widely usage of the Internet, also the number of users changed continuously at the channel that causes a variety of bandwidth availability. Such reasons make the channel status bad at the network traffic congestion. So the controller is used to eliminate this problem and avoid network traffic congestion. The controller is shown in Fig. 2. The goal of this work is the streaming of UHD (4k) video resolution over the Internet. H.265 encoder's features and parameters are used to find the optimal value of BR and PSNR for different levels. The raw video, 4K is passed through a subsampling process to produce the other levels, 1080i/p, 720p, 4CIF, CIF and QCIF, see Fig. 2. All levels are encoded with H.265 as a parallel process to make the controller selects the proper level based on the status of the channel buffer. In the steps of this design, each of the levels is worked of optimal PSNR and BR. When the channel suffers from traffic congestion at a time, the controller selects a video sequence with a level less than UHD to be transmitted over the Internet. While at the receiver terminal side, there are two strategies for video reconstruction, if the receiver device is displaying the video with UHD and the received video with a level less than UHD it should be processed by up subsampling or interpolation to retrieve the required level. But if the receiver is displaying different levels with several applications, it is updated to be suitable to that level, which is nearer to the scalability situation [15].

III. THE PROPOSED SYSTEM OF VIDEO STREAMING OVER THE INTERNET
The second part of the proposed system is how to get overall video streaming over the Internet by integrating the encoder operation with the network protocols to examine the network performance and its effects on the video streaming quality for different video test sequences. Three classes of video content (high, medium and low motion details) are the video sources to generate an appropriate study about network performance and its parameters, channel bandwidth, packet loss, latency and delay jitter. This mapping of information can be used in the source server to support the QoS to a multimedia distributed delivery system by using the transport and streaming protocols (UDP, RTP, RTCP and RTSP). One basic issue is the network path condition from server to client and impacts on video quality when it suffers from packet loss, latency, bandwidth congestion and delay jitter. This work focuses on how to transmit the video at a low-bit rate taking into account the network status. The multimedia data transmitting over the Internet is very important at the smart devices and applications, so the RTP and its accompanying RTCP have now been developed to distribute the multimedia over internet. This work suggests the use of the RTP to provide the functionality of transforming the real-time data over a network without flow control, while the control is achieved by the RTCP that is used to give feedback about the network path status to allow this system to change the video according to network condition. This work will try to achieve better scalability and performance of video streaming over the Internet with suitable feedback of QoS by this mechanism.  The test environment contains the source server for video streaming on the Sub-Net1 side and the client at the Sub-Net2 side as shown in Fig. 3. This system uses RTSP for streaming and transmitting by RTP after selecting an appropriate level of video by using RTCP according to network status. While the player requests the streaming content at the server, the server accepts that by RTSP and sends the video streaming to the client using RTP. At the server, there is a network emulator tool used to monitor the network path from server to client and emulate the network parameters conditions. With RTCP, it is used as applied in the standards that give the reports as feedback about QoS with the reflected method that broadcast the SR to all participants in a one-to-many way, while the receiver generates the RR and sends it by the one-to-one way not having ability to send with one-to-many. The sender (server) receives the (RR) packet from each receiver (client) in unicast connection (one-to-one way), so the sender is listening to these reports and collecting them and releases to each member by using a multicast group. The (RR) report has the QoS information that goes to the source server to give it the channel status that helps it to select an appropriate level of the encoded video. Generally, any recorded video is described by sequential motion activities consisting of different details, low, medium and high. In this experimentation, there are three test sequences (HoneyBee, Jockey and ReadySetGo) with different details and motion, as briefly described in the following.
• The first one is a "HoneyBee" sequence with an object of little movement with a fixed lens, the characteristic of this video is low details.
• The second one is a "Jockey" sequence with the dynamic moving scene and passing the camera to the left, the characteristic of this video is medium details.
• The last one is a "ReadySetGo" sequence with one moving scene and passing the camera to right, the characteristic of this video is high details, see

A. Methodology of Encoding Model
The encoding model that is implemented in this work is described in section 2.  of H.265 and its features can get the optimal encoded video of each video level. In this work, "ffmpeg" software is used to implement H.265 encoding over three test sequences with changing encoder parameters that directly affect video bit rate and quality. The video quality is represented by the PSNR value. The ffmpeg libraries (libx265 and libavcodec) are used to produce the compression video streaming. The H.265 encoder is applied on raw test sequences with yuv format using coder parameters (QP, CRF, GoP and REF), that are changed to get optimal video bit rate and quality.

B. Methodology of Video Streaming Over The Internet
The architecture of network streaming as shown in Fig. 3 is implementing by Graphical Network Simulation (GNS3), it can provide a virtual workstation for real network devices. The proposed scenario in this work applies to the network topologies and protocols that now existing. For installing the server device with Windows Server2012 R2, used the virtual machine Qemu template. This server includes one network adapter Ethernet type e1000 interface. Also installing the Client by Qemu VM template with Windows 10. However, the CISCO router with c7200 is installed by Dynamips IOS router template. The structure model of video streaming network content has two scenarios: Unicast and Multicast video streaming.
The video is streaming from server to client in a one-to-one way at unicast, while at multicast it's streaming from server to four clients, through the channel (router) that implement in GNS3 as shown in Fig. 6 (a) and (b). At the server, the VLC-player software uses to streaming suitable video sequences, that is prepared with six levels of spatial resolutions. The Cisco router has been configured with suitable IP addresses for two interfaces (Fastethernet 0/0 and 0/1). The interfaces of each device should be set for the unicast and multicast streaming network. Which at the multicasting, the Open Shortest Path First (OSPF) protocol should be configured to achieve reachability. Also, the IGMP group for the multicasting process is implemented on the router interface (f0/1) with IP address (224.4. From the experiments of three types of test sequences, can conclude the two scenarios: • First of all, is how to find the optimal configurations of the encoder's parameters with each video layer that gives the required PSNR (36 dB) as an initial value the system may work with it at the beginning of video compression. While the controller is incorporated with this encoder to achieve the video streaming adaptation on the available bandwidth by selecting an appropriate layer at an assigned bit rate that is compatible with the channel and keeping a good quality for the end-users in the optimal state of the network. From these arguments, one can decide what the optimal configurations of parameters are used, where the value of QP changing according to video's level, will GOP and REF fixed at default values (GOP=100 and REF=3). This scenario applies to three test sequences but reviews for video with medium details as shown in Table I. of encoded video to be transmitted over the channel. While at the smallest resolution (QCIF) can accept the PSNR more than 36 dB.

B. The Experiment Results of Video Streaming
The evaluation of network performance has two categories: (advanced and simple investigation methodology). The  Fig. 7, 8 and 9) show the results for medium details Jockey video. Fig. 7 describe the (077) Ethernet traffic sensor's results of the advanced investigation, the methodology/bandwidth availability.      For process time saving of adaptive compression, the critical parameters (QP) of the encoder were applied with four states to finding the best range of acceptable quality (32-40) dB. There are two scenarios of the encoding part, at the beginning the system work with an initial value of PSNR (36) dB, it is achieved by finding the optimal configuration of the encoder's parameters. Then changing the value of QP to get the best range of each video level of resolution as shown in Table II.
These ranges can be applied to any video with different motion details and long duration. While at the experiments of the second part of this work, get a global scope of finding the QoS map for the unicast and multicast network performance. At the unicast network, the maximum value of the total traffic for low details video is reached to 10 Kb/s, while the medium details are reached to 16 Kb/s as shown in Fig. 7 , but with the high details is reached to 80 Kb/s. Also, the range values of the jitter are from 1 to 5 for their videos as shown in Fig. 8. The values of the ping time range (12-26 msec) as shown in Fig. 9, while the maximum latency at the ReadySetGO is 28 msec, the ReadySetGo video suffers from delay more than the other two types of videos. At the multicast network, the max value of the traffic achieved to (25 Kb/s) for HoneyBee and (74 Kb/s) for the Jockey as shown in Fig. 10, while at the ReadySetGo achieved to (108 Kb/s). The values of jitter at the four clients range is  for the HoneyBee and jockey as shown in Fig. 11, while the jitter range at ReadySetGo with high motion details is (2-101) it suffers from jitter more than the other two videos. The delay for this video achieved 195 msec which is more than the other two types of video. The important difference between the results of bandwidth availability, jitter and delay is based on several reasons including the video's content categories that are affected by buffer size, encoding scheme, and video in terms of size, duration, frame rate and bit rate, and finally the user's number at the network.

VII. CONCLUSION
In this study, the measurements methodologies of the proposed system of video streaming over the Internet, there are several conclusions. The goal of this study adapted transmitting the UHD video over the Internet according to the channel status and increased the number of users. In this work, there are two topics for the problem solution, video encoding and streaming over the Internet.
• The first topic about encoders: 1) The QP is the most effective parameter, the order of QP adaptation is controller signal based on buffer state.
2) The controller incorporating to choose the suitable format from H.265 multi-layers of sequential resolution.
3) The switching between layers is based on the controller signal buffer state, which achieved scalability. 1) The experimental is tested on two networks: unicast and multicast to give an integrated map about the QoS, that applied by streaming three categories of video details on these networks.
2) Each video is monitored with a PRTG monitoring system to get the reports of results to conclude the measurement based on details of the video. From the QoS, the map can be predicted how network conditions affect video streaming over the network.
Furthermore, for optimized the system performance by applying the adaptive compression and adaptive layer selection. The results of the proposed algorithm experiment show the reduction of processing time and improvement of received video quality