|
|
Contents |
DVI, MPEG, Indeo, Ultimotion, FMV, AVI, QTW, VfW, ActionMedia II, MediaBurst, VideoIn, NTSC, Software Motion Video, S-VHS, JPEG, MJPEG, Quicktime, MCI... Ever wondered what they all mean? This document will attempt to explain them all and give you some idea of what you can do on your PC. The first thing you must understand is that different people mean very different things when they talk about motion video on a PC. If you are talking about a video that is going to be used in public to merchandise your wares, e.g. in a multimedia kiosk, then you need the best quality you can get (price permitting!). If you want to send a short clip to a friend over the Internet then you may have very low quality requirements and no budget whatsoever! (Note all prices are approximations in US dollars.) The main problem with video is that it is ENORMOUS! If you do the calculations for a full screen, full motion video with high quality colour and high quality sound: Movie Frame Size: 640 x 480 Colour Depth: 16 million colours (24-bit) Frame Rate: 30 frames per second Audio Properties: 44.1 KHz, 16-bit, stereo the total date rate is 27,648,000 (video) + 176,400 (audio) bytes per second. That is 28MB per second! Unfortunately the average PC hard disk can probably only achieve about 300KB per second which is about 90 times too slow - and CD-ROM is even slower. N.B. Please note that MB is megabytes and Mb is megabits! So, the only way you can do video on a PC is by throwing away some of the data (which may reduce the quality), compressing what you have left (which may reduce the quality) and, maybe, shrinking it even more to match the speed of your PC (which may reduce the quality!). Although the compression process can take as long as you like, when you play the movie back on your PC it has to be decompressed in real time as it is being played. If you have specialist hardware to do this on movies which were only intended to be played with that specialist hardware then you are using Hardware Motion Video - the two main classes are DVI and MPEG. The specialist hardware provides more power to do the decompression so you can have higher quality movies. MPEG and DVI typically take about 10MB per minute. If you are playing a movie without any special hardware then all the decompression has to be done with software running on your PC's processor - this can only cope with a certain level of work so, in order not to overstress it, the movie will normally be in a small window, jerky and grainy. This is Software Motion Video. Quicktime for Windows, Video for Windows and OS/2's Digital Video Player are all Software Motion Video Players. BUT... in order to be able to play back a Software Motion Video at a higher quality you can now buy hardware to assist in the decompression process and do some value-add as well to further improve the quality of what you see. This is known as Hardware-Assisted Software Motion Video! AND... in order to play MPEG on any PC people have produced software MPEG players - how well they work will depend on the power of your PC. So the lines between hardware and software motion video are rather blurred - basically if it looks really good you can be pretty sure there is some specialised hardware helping it along! CODEC - Compressor/Decompressor there are lots of different algorithms for doing the compression - the hardware or software that does it is called a CODEC. NTSC/PAL/SECAM These are television standards in different countries. NTSC (National Television Standards Committee - Never Twice the Same Colour!) is used in North and Central America and Japan. PAL (Phase Alteration Line) is used in the UK, Western Europe, the Middle East and parts of Africa and South America. SECAM (Systeme Electronic pour Couleur Avec Memoire) is used in France, Russia, and parts of Eastern Europe and Africa. If you play a SECAM video on a PAL player you will get the picture OK, but no colour. If you play an NTSC video on a PAL player you may get colour OK, but the picture will be rolling up the screen so badly that you can't see it properly! FRAME one still picture from a film. NTSC runs at 30 frames per second. PAL and SECAM run at 25 frames per second. FIELD a frame for television is made up of two "fields" - one contains the odd-numbered lines, the other contains the even-numbered lines for display on your TV. FULL MOTION VIDEO (FMV) this mean running at 25/30 frames per second - it is the motion that is "full". FULL SCREEN MOTION VIDEO this normally means 640x480. Actually this is only "full screen" if you are still living in the 80's! Most modern computers can run at much higher resolutions! SCREEN SIZE the VGA standard was 640x480 - that is what most PCs in the late 80s were using. IBM enhanced this with XGA, but the clone industry invented SVGA (in a thousand different flavours!). XGA-2 provided some of the popular SVGA formats. All IBM's current mainstream PCs are now SVGA. Another key feature of your screen's capability is how many colours it can display. The basic number is 256. Unless you have some extra video memory in your system you will not be able to display more colours. Extra memory is normally 1 or 2MB and may be either VRAM (Video RAM) or DRAM (Dynamic RAM). The number of bits required to represent each pixel is known as the COLOUR DEPTH. SVGA gives: With 1MB (COLOUR DEPTH) With 2MB 640 x 480 16 Million (24-bit) 16 Million 800 x 600 64 Thousand (16-bit) 16 Million 1024 x 768 256 (8-bit) 64 Thousand 1280 x 1024 16 (4-bit) 256 1600 x 1200 -- 256 N.B. Most laptops are currently limited to 640 x 480 with 256 colours on their internal screens - the latest IBM Thinkpad 755 CE and CD can do 64 Thousand colours. PCI/VESA LOCAL BUS in order to get the best possible performance you need to get an awful lot of data to your screen as fast a possible. The standard ISA bus is not really fast enough so many graphics accelerators now come in VESA Local Bus or PCI (PC Interconnect) versions. Neither PCI nor VESA is intended as a replacement or competitor for either ISA or MicroChannel. They are COMPLEMENTARY technologies to allow the attachment of ONE or TWO high speed devices to your PC. The IBM PC range uses VESA Local Bus on the 486 machines and PCI on the Pentium machines. REAL TIME & ASYMMETRIC COMPRESSORS the simplest way to capture a video into digital format is to plug your video tape into your capture card, play a video tape and as it plays let the capture software create a digital file as the tape plays - in real time. This limits your compression software to 1/30th second (for NTSC) to compress each frame. An Asymmetric process uses more sophisticated equipment to allow you to process the frames at your own speed and will normally give far higher quality. The catch is of course that the equipment may be very expensive and it may take hours to capture a few minutes of video. INTRAFRAME COMPRESSION (spatial compression) this means reducing the size of a single frame - all compression systems do this. The Discrete Cosine Transformation (DCT) is very popular! There are two forms of intraframe compression: lossy and lossless. If you use PKZIP or SAVERAM to compress some files and send them across the network, then when they are PKUNZIPed or LOADRAMed you expect the resulting files to be IDENTICAL to the originals. This is lossless compression. However, when you are working with images you can take advantage of the limitations of the human eye - you can actually throw quite a lot of data away and still be able to reconstruct an image which the human eye cannot easily distinguish from the original. If you are prepared to accept slight degradation in quality you can make your compressed frame even smaller! This is known as lossy compression - most systems do this. INTERFRAME COMPRESSION (temporal compression) the next step is to compare each frame with its predecessor and only keep the pixels that have changed. Normally this is done by using reference or key frames which are complete in themselves - for each frame up until the next key frame you just store the difference between the new frame and the previous key frame (key frames are called I-frames in MPEG). How effective this is depends on many factors, not least of which is the content of the movie. A "talking head" may have very small frame differences - just the lips moving. A zoom or a pan tends to make EVERY pixel change. MCI the Media Control Interface is the set of commands which allow you to interface in a standard manner between your application software and the hardware through MMPM/2 (Multimedia Presentation Manager for OS/2) or MME (Microsoft Multimedia Extensions for Windows). Support for new hardware can be added as black boxes below MMPM/2 and MME and your application code should still work. The players on you desktop all use MCI strings to communicate with the devices. VIDEO CAPTURE this is the process of taking an analogue video source (VTR, VCR, Laser Disk player, Camera, etc.), feeding the analogue signal into your capture hardware and generating a compressed digital file on your hard disk. When you are not happy with the quality of a piece of video you are playing on your PC there are TWO important questions to ask: Is my PC actually capable of playing this video well at this size? Have I got all the hardware and software I need? Are all the components big enough and fast enough? And are they set up properly? How was it captured? The second question is absolutely critical to achieving quality. The steps are: Choose suitable material - zooms, pans and flashing lights all make it harder for the PC to handle; shots of TVs or computer screens can look terrible unless you use the right synchronisation hardware on the camera; monochrome backgrounds may be boring, but they can result in higher quality digital videos. If you can affect the shooting of the original material then get a specialist to help advise the film-maker on how they can shoot material which will digitise well. Get a high-quality original. Tape standards include:
Professionally produced material will be better than stuff you copied from the TV! If you have a copy of a copy then don't waste your time! Connect it to your hardware in such a way as to provide the maximum possible information.
And don't forget that that is just the image - you need to connect the sound up as well! Asymmetric processing - which normally requires specialist hardware so that the PC can control the video source and read one frame at a time - will give you better quality than real-time capture. Make sure all the components of your hardware are capable of creating high quality movies. Get all the parameters right - your capture software will have lots of options and some of them will drastically affect the quality of the output. The last point is especially important for Software Motion Video - the movie may have been captured at low quality INTENTIONALLY as a trade against some other parameter such as the speed of the generated data stream so that it will play back on slow machines. N.B. You only need ONE weak link in the chain to ruin all your work and leave you with low quality video! You generally have much less control of the playback, but some of the players give you some options, either at set up time or as you play - check your documentation carefully and see what you can experiment with. A good example is ActionMedia II RTV - it often comes out looking rather red - all you have to do is reduce the Colour Saturation at play-back time! The big commercial breakthrough came with the IBM/Intel DVI (Digital Video Interactive) technology. IBM made this available through the ActionMedia II card. At over 1500 dollars when it first came out it was not cheap - but could produce spectacular results. Other companies produced their own versions of DVI adapters. It is important to understand that there are two very different ways of capturing DVI video: RTV Real Time Video - for less than a thousand dollars you can add a daughter board to your AMII card and capture your own video in real time. IF you get the parameters right you can capture quite reasonable video with the sample program provided with AMII. BUT... RTV only really works well at quarter screen size. Above that you have to drop in to AMII's low resolution mode and the quality suffers badly. There does seem to be some synergy with full-screen size - it seems better there than at some intermediate size. PLV Production Level Video - if you send a high quality video tape (e.g. Betacam SP format) to a firm in Paris (if you are in Europe) then they will use an Asymmetric process to capture a much higher quality video for you - it will take days (maybe weeks?) and cost you money (maybe 100 dollars per minute - I haven't checked lately). You can play PLV at full screen size with excellent quality. With an ActionMedia II card in your PC you can play either RTV or PLV movies. ActionMedia II is fully supported under OS/2 and Windows (though you may need to run a utility to reformat the file slightly if you move between OS/2 and Windows). AMII uses a Y-cable to link the output of the AMII card to the output from your SVGA card and send the two signals to your screen. The AMII signal will only show through areas you have blacked out. This has lots of interesting side effects: AMII is not limited to the number of colours your graphics card can see; AMII works at a percentage of screen size, not at a number of pixels - it doesn't know how many you are showing - so "full screen" means "full screen" regardless of the resolution; you can't use something like PMCAMERA to grab a screen image with a video frame since the SVGA adapter has no idea what is being shown; you can do artistic effects with parts of images superimposed on the video. Unfortunately ActionMedia II's days may be numbered. Intel have stopped all development. They are now concentrating on their Indeo Software Motion Video - indeed the last release of DVI support they provided was to allow the use of DVI cards to accelerate Indeo. IBM has contracted another supplier to provide a new, albeit potentially limited, supply of adapters and is still fully supporting ActionMedia II. For 1995 DVI is undoubtedly a viable solution to many digital video requirements. Through 1995 and 1996 we will see MPEG maturing to take its place in many areas. IBM is watching the market carefully to provide the best future solutions for our customers. Most of the alternatives to DVI can only provide high quality quarter screen video by using expensive asymmetric equipment for the video capture process. One of the strengths of AMII is the very cheap real time capture capability. The improvements in Software Motion Video and the rapid acceptance of MPEG as a Hardware Motion Video standard will continue to have a significant effect on ActionMedia II's future. The current AMII card price is around 1250 dollars, plus 400 dollars for the capture option. There are three versions of MPEG: MPEG-1 is an international standard providing VHS quality video at around 1.5Mbps. This speed was chosen specifically to allow transmission on POTS (Plain Old Telephone System! - i.e. the copper wire that goes into your home) and to allow playback from single-speed CD-ROM drives. It is already used for both those purposes: you can buy feature films on CD now to play on your home-PC; many digital Interactive TV trials use MPEG. MPEG-2 is an international standard (as of Dec 1994) providing broadcast transmission quality video at from 3 to 10 Mbps. The standard also provides transport protocols of relevance to using MPEG-2 across networks. MPEG-4 is a new standard currently under discussion. The original focus was on low bandwidth technology; now it is focusing on Content Based Interactivity, Compression and Universal Access. Completion is scheduled for 1998, by which time they might have decided exactly what it is! MPEG stands for Motion Picture Experts Group. The compression algorithms are very powerful and normally require hardware to perform the decompression. The rest of this section is mainly about MPEG-1. There are programs available to provide software decompression of MPEG files, but the quality of playback may not be very high unless you have some hardware in you PC to help with the process. Modern PCs often have multiple specialist processors. As well as the 486, Pentium or PowerPC which is your main processor you may have a DSP (Digital Signal Processor such as MWave) or a Graphics Accelerator (either built in or on an add-on card). Clever programs are appearing which use these specialist processors to boost the performance of software MPEG decompression resulting in a higher quality of video playback. The simplest way to achieve high quality playback is to add an MPEG Decoder Card to your PC. One of the early leaders in the MPEG playback market was the Reel Magic card from Sigma Designs. The current Real (they changed the spelling!) Magic Lite card is only about 350 dollars. OS/2 and Windows both support MPEG playback using the Real Magic card. There are at least three dozen other MPEG Decoder Cards on the market already, ranging in price from 250 to 1500 dollars, plus some graphic accelerator cards which incorporate MPEG decompression. Most of the cards on the market are aimed at MPEG-1 at the moment. They all have Windows drivers. Check carefully for driver availability if you want to run under OS/2 Warp. MPEG-1 is intended for image resolutions of approximately 360 x 240 - the exact size will depend on whether you are using SIF (Standard Interchange Format) or CIF (Common Interchange Format) and whether you are using NTSC or PAL as your source - a typical PAL CIF video would normally be 352 x 288. This will give you slightly bigger than quarter screen if you are using 640 x 480. How good it will look at other sizes and at full screen size will depend on how good your MPEG Decoder Card is - one that interpolates the colours of extra pixels will look better than one which just duplicates the existing pixels. So with a relatively cheap MPEG Decoder (decompression) Card in your PC you can play high quality MPEG-1 movies. However, if you want to capture (compress) your own MPEG videos it is rather more expensive! MPEG compression is best done as a multi-pass asymmetric process on powerful equipment such as a IBM PVS (Power Visualisation System). It will cost you 50 to 100 dollars per minute of video. IBM provides this facility on a commercial basis in locations such as Hursley, UK - contact Jason McMahon on 01962-815123 (UK) and Atlanta, USA - contact Jeff Ramminger on (404) 835-9366 (USA). It is important to note that at this stage MPEG Compression is more of an art than a science. For example, it is possible to apply filters during the compression process to improve the quality, and you can tune the process depending on what the target device (PC or TV) is and which MPEG Decoder Card will be used - a digital movie created by an experienced professional with an extensive production suite will be better than anything you can create automatically. You CAN buy MPEG compression adapters for your PC - the best known one is probably the Optibase one at around 20,000 dollars. You will NOT get the same quality from this as from a full PVS system, YOU have to decide what quality YOU need for your application. New offerings are appearing on the market all the time - check carefully for what quality they can achieve and also what size of image they can generate at 30 frames per second. The latest low-cost offering is the RealMagic Producer from Sigma Designs - for less than 5000 dollars you can now produce your own MPEG videos on your PC - yes, it works and is better than most Software Motion Video - no, don't expect the quality you will see on a Video-CD! To make it easier to control MPEG-1 playback from CD a number of companies have developed a VideoCD ("White Book") standard to add data-interchange and VCR-like controls to MPEG-1. On March 28th 1995 IBM introduced the world's first commercially available MPEG-2 Digital Video Single-Chip Encoder, plus an enhanced MPEG-2 Decoder chip aimed at the consumer electronics market. Both chips are compatible with MPEG-1. Watch this space! MPEG and DVI are nice and simple compared with SWMV! MPEG cards play MPEG files. ActionMedia II cards play DVI files. Not so with SWMV! There are countless CODECs available - some are achieving de facto standards status, but new proprietary ones are emerging all the time in conjunction with new capture adapters from various companies. You have to use the CODEC to create some compressed data and then store it in a file - and there is more than one format for that. Then you need a Player to play the video on your PC and none of the players understand all the file formats or CODECs! Some of the popular CODECs are:
The main file formats are:
The main players are:
A lot of videos are captured on Macs and stored in .MOV files - they cannot then be played on OS/2 without conversion to .AVI format. QTW can play Cinepak and various other Apple formats. VfW can play all of the ones listed above (including Ultimotion by using an as-is decompressor available on the Internet). MMPM/2 can play Indeo and Ultimotion (and there is a MS Video 1 decompressor available on the Internet). The biggest benefit of SWMV is that ANYONE can play it on ANY modern PC - you don't have to buy extra hardware. (Actually, you will probably want a sound card if your PC doesn't have audio capabilities built in!) The biggest problem with SWMV is the quality! You have all probably seen jerky, grainy video clips playing in tiny windows - would you be prepared to sell that to your customer as a level of quality he should be happy with for corporate training material or selling his products to the public? Well, that's not quite fair actually - it is possible to get quite good quality. QuickTime for Windows with the Cinepak CODEC is one of the best quality combinations as used on the Windows version of the IBM Interactive Annual Report 1994; the OS/2 version uses a derivative of Ultimotion to achieve similar quality. (Hopefully this new Ultimotion will be made available soon.) The quality is mainly dependent on how the video was captured. As discussed above, there are many factors which can affect the quality of the video capture process. DVI and MPEG are normally trying to achieve full or quarter screen, with high quality colour, high quality sound and high frame rates - they can afford to do this because they KNOW you have special hardware to help with the playback. SWMV, however, is limited by the capabilities of the slowest machine you plan to play it back on. As well as choosing a CODEC the specialist who is capturing the video has to decide what data rate he is going to limit himself to and choose a frame rate, frame size, colour depth, and sound quality which will fit into that data rate. He may actually capture at a higher data rate and then use a software tool to reduce that rate, smoothing it out as it does so. The quality of the tool will obviously affect the end-result as well. If the specialist is inexperienced or in a hurry he may accept the quality of whatever he first creates. If he cares enough and has enough time he may redigitise with different parameters, possibly also playing with the colour, brightness and saturation controls to try and achieve a higher quality. A typical SWMV may be designed to be: 240 x 180 pixels 24-bit (16 million) colour 15 frames per second 11.025 KHz, 8-bit, mono sound If you are playing a movie like this you may notice the following effects:
There are over 50 Video Capture Cards available for the PC, ranging in price from 199 to 20,000 dollars. They vary considerably in their ability to handle all the above parameters. All the factors mentioned above under VIDEO CAPTURE QUALITY and below under HIDDEN VIDEO CAPTURE COSTS can also affect the quality of your videos! OS/2 Warp includes the Video IN Recorder software to allow you to capture SWMV - you will still need a Video Capture Card. There are a number of other powerful video capture packages on the market, many of them come free with suitable Video Capture Cards. IF you have a powerful machine, good software, good technique and a high quality video source then it is possible to get some pretty impressive stuff playing on your screen! BUT what if that isn't good enough? That brings us to Hardware Assisted Software Motion Video! You may have seen some very impressive demos of what you were told was SWMV, oh, well, yes, they may have mentioned that there was a Graphics Accelerator helping it along. The Graphics Accelerator is actually a piece of HARDWARE - so we are moving away from pure SOFTWARE Motion Video! Accelerators normally provide three main functions: Provide hardware decompression so that you can achieve higher frame rates. Get the data to the screen faster so that you can have bigger frame sizes (especially if they use VESA Local Bus or PCI). Interpolate i.e. make intelligent decisions about how to colour extra pixels when you enlarge the image so that the picture looks smoother. There are countless Graphics Accelerators on the market offering varying capabilities. IBM has three different offerings at the moment: Thinkpad 755CD This has a built in Video Accelerator Chip (an ASCII V7320) providing video playback with 64K colours at 640 x 480. It supports Indeo 3.1, CinePak and MS Video 1 at 30 frames per second under VfW. PS/2 m77s - Mediaburst (not announced in all countries) This has built in VideoLogic chips giving similar capability to VideoLogic's 928Movie Card. Support includes Indeo, Cinepak and Video 1 under VfW and OS/2 Warp. PC 700 Series Optional High Performance Video Adapter (actually a Diamond Viper Pro). So, if you have a really high quality video that was captured at 640 x 480 with 30 frames per second and 24-bit colour, and you can read it from your hard disk fast enough, then you can get superb results from using an Accelerator. I still haven't seen a sample of that quality! Catch-22 is of course that most SWMV clips are captured at much lower quality than that (most Video Capture Cards can't achieve it anyway!), so, even though they may look better with the Accelerator than without, they will not be as good as DVI-PLV or MPEG! And, if you do manage to get clips of that quality, they may look great with your Accelerator, but probably won't play at all well on the average PC without an Accelerator - thus losing one of the benefits of SWMV: its universal playback capability! With current technology you can't have everything. You must decide what quality you require and what type of machines you need to be able to play it on, then make some decisions. You now know about the MPEG-1 and MPEG-2 International Standards. You may also have heard about JPEG (Joint Picture Expert Group) which is an International Standard for still images. JPEG provides a range of powerful lossy and lossless algorithms which are gaining rapid acceptance throughout the industry. But many people are confused when they hear of MJPEG or Motion JPEG. You just take each frame of the movie and treat it as a still, compress it with JPEG, then string them together at the appropriate number of frames per second and you have a movie! Benefits:
Disadvantages:
There are a dozen or so cards which will do MJPEG compression for you on your PC - but you really need to be able to read one frame at a time from your video source in order to capture all the information you are going to compress and store, and a very fast disk subsystem to put the data out to. MJPEG has found good acceptance in some areas of the market. It is widely used at the high end, often as a way of creating a digital source for the generation of multiple SWMV format files. (Just for completeness I should probably mention JBIG (Joint Bi-level Image Experts Group) and MHEG (Multimedia Hierarchical Expert Group)! JBIG is a new technique for bi-level or black/white image compression that will improve the compression of facsimile images. MHEG is a Draft International Standard for the encoding of multimedia presentations - so if you want to distribute some video, audio, stills, text, etc. with which the end-user can interact and which are synchronised with each other, then MHEG may be the way you do it in the future. Now practise saying "M-HEG" and "M-PEG" so that people can tell which one you said!) One of the commonest questions asked is "Can I convert from xxx format to yyy format?". The simple answer is "Yes" in most cases - all it takes is an intelligent conversion program that understands both formats. Unfortunately a fuller answer is probably "No, because the quality will suffer." The problem is that the CODECs work best when given as much information as possible - by using a compressed video file as input you are automatically restricting the amount of data that the CODEC has available to work with. The result is usually disappointing. Going from one of the higher quality formats to one of the lower quality formats may give you acceptable output - but it will never be as good as starting again from the original source material. So the real answer is "It depends on whether the resulting quality will meet your requirements (and then you have to find the software!)." |