B.C. Smith, "Implementation Techniques for Continuous Media Systems and Applications," Ph.D. Dissertation, Computer Science Division - EECS, U.C. Berkeley, December 1994, ftp://mmftp. cs.berkeley.edu/pub/multimedia/papers/ bsmith-thesis.ps.gz.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....Mash. Mash Mash [49] is a toolkit for building remote collaborations and streaming media applications. It was built by merging the functionalities and design ideas from the MBone tools, the VuSystem from M.I. T [44] and the Continuous Media Toolkit (CMT) from University of California, Berkeley [71]. Mash is built using C and an object oriented version of Tcl, called OTcl [80] It consists of a set of 13 reusable and flexible components that can be composed to perform a function. Since Mash abstracts away many low level details, such as communication with video capturing devices and the ....

B. C. Smith. Implementation Techniques for Continuous Media Systems and Applications. PhD thesis, Department of Computer Science, University of California, Berkeley, December 1994.

....hostname or IP address = index f ilename where server hostname or IP address is replaced by the hostname of the video server or its IP address and index f ilename is replaced by the name of the index le associated with the targeted video le. Other video players also use similar formats [13, 14]. ODBC Database Server: The database server maintains descriptions and titles of the videos in its database. VideoCenter accepts keywords from the user and queries the database server to nd candidate videos whose title or description contains the keywords. Video Server: The video server ....

....sent to the server. The feedback period is set to twice the round trip time or 200 milliseconds if the round trip time is too small (less than 200 milliseconds) Feedback mechanisms are widely used for several purposes such as to synchronize data transmission and presentation of media streams [22, 23, 13, 24, 25] or to reduce the number of loss of packets. We will discuss the feedback mechanism used in VideoCenter in more detail later. 2PSMFilter performs the following tasks according to the server s response: If the server responds with an error message, the lter noti es the application to report ....

B. C. Smith. Implementation techniques for continuous media systems and applications. In Ph.D. Dissertation, University of California, Berkeley, September 1994.

....copies for videos that are deemed less important. More research is needed on this topic. Finally, many people are working on network protocols for delivering video [11] While real time protocols like RTP [12] and RTIP [13] will eventually be used everywhere, until then best effort protocols [14] will still be useful particularly when coupled with reasonably adaptation by client playback software [15] 5. ACKNOWLEDGEMENTS Many people have contributed to the design and implementation of BDVS including Charlie Destefano, Charlie Eads, Craig Federighi, Peter Liu and Brian Smith. This ....

....Engineering Task Force, Audio Video Transport Working Group, July 1994. 13] A. Banerjea, E. Knightly, F. Templin, and H. Zhang, Experiments with the Tenet Real Time Protocol Suite on the Sequoia 2000 Wide Area Network , Proc. ACM Multimedia 94, San Francisco, CA, pp. 183 192, October 1994. [14] B.C. Smith, Implementation Techniques for Continuous Media Systems and Applications , Ph.D. Thesis, University of California at Berkeley, Dec. 1994. 15] L.A. Rowe and B.C. Smith, A Continuous Media Player , Network and Operating Systems Support for Digital Audio and Video, La Jolla, CA, ....

[Article contains additional citation context not shown here]

B.C. Smith, "Implementation Techniques for Continuous Media Systems and Applications", Ph.D. Thesis, University of California at Berkeley, Dec. 1994.

....and lower cost. We are familiar with three research multimedia servers comparable to ours. Freedman and DeWitt [8] describe simulations of a video ondemand system in which clients request data as their buffers empty rather than having the server deliver the data at a pre determined rate. Smith [14] focuses on coding and protocol issues rather than on the architecture and scalability of the server itself. Gelman et al. 9] describes a multimedia server built from custom hardware. Their system includes an end to end network distribution system for video on demand that uses buffering in ....

B. Smith. Implementation techniques for continuous media systems and applications. Technical Report UCB/CSD 94/845, Computer Science Division---EECS, U.C. Berkeley, December 1994.

....Abstract The Continuous Media Toolkit (CMT) 1] 2] is a flexible toolkit which facilitates development of local and distributed continuous media applications. Data transfer across a computer network is provided on a connectionless, best effort basis using a network protocol called Cyclic UDP [3]. A second set of network protocols, called Tenet Suite 1 [4] 5] has been designed to provide a simplex, unicast, connection oriented service to realtime traffic in a packet switched internetwork, with guaranteed performance in terms of data throughput, end to end delay, delay jitter, and loss ....

....image from the shared memory segment. Finally, when the MjpegPlay object is destroyed, commands are passed to jvdriver to relinquish the shared memory segments and to the other processes to detach from it. 4. New CMT objects for Tenet Suite 1 This section takes a closer look at Cyclic UDP [3], the set of network protocols already supported by CMT. The functional requirements of the toolkit s network objects are discussed, and consideration given to those of new objects that use the protocols of Tenet Suite 1. The implementation of these new objects is described. Details are given on ....

B. C. Smith, Implementation Techniques for Continuous Media Systems and Applications, Ph.D. thesis in Computer Science, University of California at Berkeley, (September 1994).

....Motion JPEG, H.261 H.263, and MPEG are the most common video formats in use today. These formats are very similar they use the DCT, quantization, and entropy coding with intra and inter frame optimizations. We will capitalize on available compressed domain processing methods wherever possible [14]. RTP is the standard networking protocol used by IV applications [12] This protocol is used for input and output video streams. The third major design goal is to use commodity hardware. The system should operate on any set of networked, general purpose processors. This goal requires the system ....

B. C. Smith. Implementation techniques for continuous media systems and applications. PhD thesis, University of California, Berkeley : Computer Science Division, 1994.

....adjust data delivery to eliminate the network congestion. This assumes that the client can still decode the residual portions of the stream and that the server can intelligently determine what to send and what to discard, while not reducing disk load. A solution based on this principle is given in [24], where the client and server co operate on defining the order of the units to be delivered to the client and the server continues to send complete presentation units. In the CMFS, the client application handles the degradation of the quality so that the server does not need to perform additional ....

B. C. Smith, Implementation Techniques for Continuous Media Systems and Applications. PhD thesis, University of California, Berkeley, CA, 1994.

....are resident in the media player) It is important to note that since code fragments are only sent the first time the command is played, problems can occur if the all the code fragments of a needed command do not arrive at the media player. Since CMT sends data using a semi reliable protocol[11], this will rarely happen, but should the media player be told to play a command for which it has not received the necessary code fragments, the media player can immediately request the data to be sent through a guaranteed channel using an RPC call to the media server. Thus the global state ....

Smith, B. C. Implementation Techniques for Continuous Media Systems and Applications. PhD thesis. Department of Electrical Engineering and Computer Science, University of California at Berkeley (1994).

....the B frames. Error recovery Lost data packets can be ignored, retransmitted, or recovered by a Forward Error Correction (FEC) scheme. The filter by Yeadon et al. 15] and OGI s player [5] ignore lost packets. The Continuous Media Player [11] pursues the second strategy by employing Cyclic UDP [14], which retransmits lost high priority data (i.e. I frames in the case of MPEG video) to give them a better chance to reach the destination. In the Vosaic player [7] and in Columbia s VoD testbed [6] a client can demand retransmission of a lost frame. FEC is applied by Nonnenmacher et al. 12] ....

B.C. Smith. Implementation Techniques for Continuous Media Systems and Applications. PhD thesis, University of California at Berkeley, 1994.

....protocol in such aspects as data sequencing, error handling, and flow control. This approach is effective because it takes advantage of the basic services provided by connectionless protocols. In fact, many protocols were designed in this way. They include: TCP, RTP [SCFJ96] cyclic UDP [Smi94] and the protocol used in Raid distributed database system [BR89] Reliability is another issue in choosing or designing a transport protocol for video applications. Two causes contribute to the packet loss and error: 1) unreliability of underlying transmission media, and (2) network congestion ....

Brian C. Smith. Implementation Techniques for Continuous Media Systems and Applications. PhD thesis, University of California at Berkeley, 1994.

....Subsequent clients join the conference by adding themselves to the multicast group, and by each client establishing one additional unicast connection back to the server. Once the connections have been established, the server mmaps the file to memory and gradually reads video frames from disk [1]. These video frames are divided into chunks, and each chunk is sent individually. Internet traffic is transmitted using IP multicast datagrams, while native mode ATM traffic is sent using AAL5 frames over a multipoint virtual circuit. The size of a chunk can be a variable number of bytes, but is ....

B. Smith, "Implementation techniques for continuous media systems and applications." PhD Thesis, Computer Science, UC Berkeley, 1994.

....decreases storage and network costs, it increases processing cost because the data must be decompressed first. The overhead of decompression is enormous: today s sophisticated compression algorithms, such as JPEG or MPEG, require between 150 and 300 instructions per pixel for decompression [1]. This corresponds to a rate of 2.7 billion instructions for each second of NTSC quality video processed. Furthermore, the data must often be compressed after processing, incurring additional overhead. One way to circumvent these problems is to process the video data in compressed form. This ....

....using t. In other words, if H=tF and H =t F, then . Since t is sparse and the input vectors F are sparse, the resulting computation can be implemented efficiently. Two properties, one of t and one of the input vectors, make condensation possible. 1. Most elements of t have small absolute values [1]. For example, 90 of the elements in shrink by 2 have an absolute value less than 0.05. You can see this by examining figure 4. 2. The input vectors F are sparse, and non zero values are typically small integers. This property is expected, since the DCT concentrates the energy of the image into ....

[Article contains additional citation context not shown here]

B. C. Smith, Implementation Techniques for Continuous Media Systems and Applications, Ph. D. Dissertation, University of California, Berkeley, CA, September 1994.

.... motionJPEG video, how fast a workstation is needed to process the data in real time Measurements of a public domain JPEG codec [16] show that when a general purpose CPU is used for JPEG compression and decompression, approximately 300 instructions are required to compress or decompress each pixel [23]. If the image processing to be performed requires, say, 100 instructions per pixel, approximately 700 instructions are executed for each pixel. Since a 640 by 480 video stream at 30 frames per second requires processing 10 million pixels per second, a general purpose CPU must execute about seven ....

....shows the percentage of coefficients with values less than the corresponding X coordinate. 4. Measurements over a several thousand compressed images indicate that, on average, 90 of the elements in a run length encoded vector are zero when compressed using the default JPEG quantization tables [23]. H wzl F xyk H tF = H t F = Page 15 Notice that for these transformations t has many elements with small absolute values. For example, 90 of the coefficients in the shrink by 2 LGDVE have an absolute value less than 0.05. This graph illustrates a property that we hypothesize is ....

[Article contains additional citation context not shown here]

B. C. Smith, Implementation Techniques for Continuous Media Systems and Applications, Ph. D. Dissertation, University of California at Berkeley, Berkeley, CA, September 1994.

....to develop prioritized transmission schemes. For example, Priority Encoded Transmission (PET) 1] uses a variation on forward error correction where redundant data is introduced to compensate for data loss. Priority is encoded by associating high redundancy with more important frames. Cyclic UDP [6] uses a retransmission scheme that allows for prioritized delivery. Reservation networks such as Tenet [3] allow users to reserve bandwidth in the network, with statistical guarantees on the transmission properties of the packets. Prioritized delivery can be obtained using several channels. For ....

....of unplayable frames. The second algorithm does not assume known bandwidth, but assumes that the bandwidth is uniformly distributed in an interval with known endpoints. This assumption is realistic, since there exist techniques to measure the channel bandwidth at any instant of time quite reliably [6]. The algorithm computes in time O(pmn 3 ) an optimal order to transmit the frames, where p is the size of the interval. The order is optimal in the sense that it minimizes the expected maximum interval of unplayable frames. We remark that the obvious greedy algorithm (in each step, choose the ....

B. C. Smith, Implementation Techniques for Continuous Media Systems and Applications, PhD thesis, University of California at Berkeley, 1994.

No context found.

B.C. Smith, "Implementation Techniques for Continuous Media Systems and Applications," Ph.D. Dissertation, Computer Science Division - EECS, U.C. Berkeley, December 1994, ftp://mmftp. cs.berkeley.edu/pub/multimedia/papers/ bsmith-thesis.ps.gz.

No context found.

B. C. Smith, Implementation Techniques for Continuous Media Systems and Applications, Ph.D. Dissertation, Univ. California, Berkeley, 1994.

No context found.

B.C. Smith. Implementation techniques for continuous media systems and applications. PhD thesis, University of California, Berkeley : Computer Science Division, 1994. 164

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC