BROWSE TOPICS
RESOURCESABOUT THIS SITE |
Video NotesAITopics > Project Notes > Technical Information > Video Notes This page contains notes about video clips, DVDs, video software, etc. AAAI Video Archive AssumptionsWe can expect to receive VHS, U-matic, or Betamax tapes in the NTSC, PAL or SECAM formats. We may also receive film. We estimate the archive will contain approximately 1,000 hours of video. Video Notes1. Uncompressed VideoUncompressed video is sometimes called "4:4:4." The Luma signal and the two Chroma (color difference) signals are all sampled at each pixel; hence 3 samples/pixel. One hour of SD (Standard Definition) television, captured without compression, results in ~140 GB (from McDonough). A standard NTSC video frame has an active display area of 720 x 480 pixels; i.e., 345,600 pixels. If three channels (one luma and two color difference signals) are each captured with 10 bit precision (30 bits per pixel), then a single frame requires 10,368,000 bits. There are 29.97 frames per second in NTSC video, so with full three channel sampling, the resulting digital stream consumes ~311 Mb/sec. Thus, storing an hour of video requires ~130 GB. For 8 bit precision (24 bits per pixel): ~104 GB. 1,000 hours of raw, uncompressed video would necessitate 254 TB, assuming 100% redundancy. As an example, one can purchase 750 GB Iomega disks for under $220 (October, 2007). 340 disks would meet the need at a cost of ~$75K. For 8 bit precision: 272 disks, ~$60K. Uncompressed video may be captured losslessly in MJPEG-2000, the new digital cinema standard (see below). One can expect compression ratios of 2-3 (Pearson and Gill). Assuming 8 bit precision (24 bits/pixel), ~35 GB/hr is a workable assumption for storage (from Stanford Digital Library Systems & Services spreadsheet, Doom9 Forum on DVD conversion, DVD-Video Disc Workshop [section entitled "MPEG-2 Storage Capacities"] and other sources. For 1,000 hours of video (100% redundancy): ~68 TB, 92 disks, ~$20K (October, 2007). Cautionary Notes
Key Conclusion: Today, the cost and physical space required for uncompressed video are too large. It is more practical to capture uncompressed video only for materials that are in such bad shape that we may only be able to perform the capture process once. Otherwise, we can retain the original videotape or film and capture compressed video. With this approach, we would only need to purchase a few hard drives. We will only know how many once we know the state of the received videos.
2. Compressed Video & SubsamplingThe compression standard for professional video equipment is "4:2:2" (aka YUY2, YUV 4:2:2 or YCbCr 4:2:2). The luma signal is sampled at every pixel, while the two chroma (color difference) signals are sampled at every other pixel. Hence 2 samples/pixel (1 for Y, 0.5 for Cb, and 0.5 for Cr). This reduces the bandwidth and storage by one-third with little to no visual difference. DVDs and MPEG-2 MPEG-2, the standard format for DVDs, uses "4:2:0" subsampling. The luma is sampled at every pixel. The two chroma signals are subsampled at a factor of 2 both horizontally and vertically. Hence 1.5 samples/pixel (1 for Y, 0.25 for Cb, and 0.25 for Cr). Assuming the source signal precision is 24 bits/pixel [3x8 bits/pixel for YCbCr], the result is 12 bits/pixel. The bit rate and storage are thus cut in half. MPEG-2 uses temporal as well as spatial redundancy. The net effect is to reduce the bit rate for NTSC from ~311 Mb/sec (as noted above) to ~9.8 Mb/sec (peak) for DVD quality. Hence ~4.7 GB is required for one hour of DVD video. Sample Hollywood Films: Bridget Jones's Diary - 6.51 GB; Bridget Jones: The Edge of Reason - 7.81 GB; Bourne Supremacy - 7.93 GB. See also DVD Demystified Glossary. For MPEG-2 compressed capture, 1,000 hours of video would necessitate ~9 TB (100% redundancy), ~13 disks, ~$3K (October, 2007). Video Clip Notes1. StanfordA starting point could be to use the two-clip approach used by the Stanford Libraries for the R. Buckminster Fuller Collection: low-res (analogous to thumbnails – that may also suffice for some individuals (e.g., viewing on handheld devices) and hi-res (that can be shown in a classroom). Check with some professors with video experience to determine the minimum requirements.
"Two 8-bit MPEG-2 files were created for each original video or film: a master file with color compression levels at 4:2:2 and a constant data rate of 15 Mbps, and a derivative file with color compression levels at 4:2:0 and a variable bit rate of 4 Mbps. Both files have a window size of 720 x 480 pixels and a frame rate of 29.97 fps. The audio data is stored as MPEG-1 Layer 2 with a data rate of 384 Kbps and sampling rate of 48.1 KHz."
"The streamed files accessible through this web site ... are in Windows Media format with a window size of 240 x 180 pixels, and two data rates are available (56 Kbps and 300 Kbps)."
Stanford uses asx files and Microsoft Media Server (MMS) protocol. The MMS protocol is used by Windows Media Services to stream data between the Windows Media Player and Windows Media Server over TCP and UDP. RGS: As I understand it, MMS is being replaced by Real Time Streaming Protocol (RTSP). <asx version = "3.0">
<title>Buckminster Fuller Project</title>
<author>Stanford Libraries</author>
copyright>Stanford University</copyright>
<entry>
<ref href="mms://vodwins.stanford.edu/bfuller/m1090/s17/ss5/0/3c.wmv&"/>
</entry>
</asx>
2. EVIA Digital Archive(Ethnomusicological Video for Instruction and Analysis). Indiana University. "An MPEG-2 copy at DVD quality (approximately 7 Mbps) will be created so that users have a high quality copy for viewing and projecting on large screens in a classroom. A 1 Mbps MPEG-1 version will be created for general workstation use, and a 300 Kbps will be created for use on low quality networks and for thumbnail versions of the video. These files will be generated on the fly using Telestream's FlipFactory, which will allow us to better serve end-user configurations as well as reduce the amount of storage space required."
3. Additional NotesA/V files are typically streamed from a Web server. The use clicks on "Play" in the client (e.g., Windows Media Player) and the player progressively downloads the AV file to the user's machine. Playing starts as soon as enough of the file has been downloaded to fill a buffer on the client machine. However, with WMP, if the user wants to view only the last few minutes of the video, there will be a lengthy delay before playing starts, while essentially all of the A/V is downloaded (or worse, the A/V starts from the beginning). The most practical way to reduce this latency is to chop the original A/V up into many pieces, then use an asx file to string them together. With this approach, when the user asks for the last few minutes, it is possible to write code to first select the correct piece of the overall A/V, then stream to the correct time within that piece. Unfortunately, WMP does not support any other approach at present (e.g., specifying a byte count in the original A/V file at which playing is to start). This is part of a range of problems arising from WMP's lack of support for the HTTP 1.1 Range header. Contact RGS for additional information. An alternate approach is to use Flash Video (FLV). This technology does permit streaming from a set start time without the need to chop up the video file. See Google Video for examples. 4. YouTube and Google RecommendationsYouTube: "... save the video in a format that YouTube can accept in order to upload. Unless you're a professional video producer, we recommend that you save your videos as either QuickTime .MOV, Windows .AVI, or .MPG files—these are the most common formats and they work well within our system. We specifically recommend the MPEG4 (Divx, Xvid) format at 320x240 resolution with MP3 audio." Google Video: "... We accept many popular formats, such as AVI, ASF, QuickTime, Windows Media and MPEG formats. For example, .avi, .asf, .mov, .wmv, .mpg, .mpeg, .mp4, .ra, .ram, and .mod are all acceptable file extensions. Please note that we currently do not accept Flash files. Specific video codecs we accept include H.264, H.263, MPEG 1/2/4 and motion JPEG. ... The frame rate should be above 12 frames per second. The bit rate should be above 260 Kbps. ... we recommend the specifications below for maximum quality and reasonable file size:
Streaming video files are viewed as Flash Video (FLV) on Google Video with Adobe Flash Player 7.0+. ResourcesSuppliers
Articles
|
