GE/RCA: 72 MINUTES OF FULL MOTION VIDEO ON COMPACT DISK

(Editor’s note: This article appeared in the April 1987 issue of Data Storage Report and described the first efforts to squeeze full motion video onto an audio compact disk.)

RCA David Sarnoff Laboratories has developed a new breakthrough technology that makes it possible to store 72 minutes of full-motion video on a compact disk. Up to now the most anyone could stores was a few minutes. The dramatic increase in storage capacity comes from using video compression techniques to greatly reduced the amount of data needed to represent each frame of video being recorded.

"Three years ago, the labs began a project to develop an all digital system combining computer graphics with television video for the business, education, and consumer market," says Arthur Kaiman, Director of digital products research at RCA Laboratories. A major requirement was that the system be digital rather than analog. A digital system would offer the user the maximum of control and interactivity. The work began with an approach conceived by Dr. Lawrence Ryan of RCA Labs.

"We believe that the new capabilities of this technology approach has itself created a new technology, one that brings together digital technology, video technology and interactive technology," Mr. Kaiman explains. RCA calls this new technology DVI Digital Video Interactive. This contrasts with the CD-I (compact disk interactive) technology introduced by Sony and Philips last year. At the Microsoft Corp. second annual CD ROM conference held in early March in Seattle Washington, RCA introduced only the capabilities of DVI. Mr. Kaiman says that the product and business development plans will be forthcoming over the next several months.

Getting to Know DVI

DVI can contain a maximum of 72 minutes of digital full-motion. This contrasts with CD-I which permits only 30 seconds of digital full-motion video. Because the video is digital, it can be manipulated just as a screen of alphanumeric data can be manipulated. In addition, DVI contains high powered motion graphics--a 3D image can be rotated for example, multistream high quality audio, and applications programs all on a single standard compact disk medium.

To achieve television quality video, to store a single video frame in digital form requires 600 kbytes. With the CD ROM maximum storage capacity of 500 Mbytes, means that it is possible to achieve only 25 to 30 seconds of full-motion video. Moreover, it will require one hour to play this 25 second of video because of the limited bandwidth of the CD ROM drive. Thus, though video can be stored on CD ROM, not much can be stored, nor can it be played back in full-motion.

GE/RCA Labs got around the problem by compressing the data since any video frame contains a great deal of redundant data. To achieve over an hour of full-motion video requires compressing the data at least by two orders of magnatude.

Moreover, a CD ROM player has a maximum transfer rate of 150 kbytes/second. To achieve a full-motion video requires that the storage medium has to provide the display 30 frames of data per second. Dividing 30 into 150 kbytes specifies that a frame of data stored on the CD ROM has to be compressed to 5 kbytes. With this compression, the drive can supply the frames to the system at a rate that produces full-motion video yet enables a large number of video frames to be contained on the disk. GE/RCA stores each video frames in less than 5 kbytes so that it is possible to store audio, graphics, and program information on the disk as well. The DVI technology uses the standard CD ROM disk and uses the same player.

On-line/Off-line

The approach GE/RCA consists of two parts, off-line and on-line. Offline the approach runs the video images through a compression algorithm, which is an integral part of the DVI system. The compression algorithm typically runs on a VAX 11/785 computer. The algorithm is optimized for compressing video. The larger computer is needed if the applications requires a large amount of video to be compressed. On the VAX, it takes 2 minutes to compress a frame, but we know how to compress at a rate of two to three frames a second.

If the application needs only small amounts of video, the compression can be done on smaller computers. The smaller computers will take much longer to do the compression than if the task were run on a larger machine. "We are investigating getting the price of the compression equipment down to something reasonable. One alternative is to use parallel processors in doing the compression.

The compressed data is then pressed onto the CD ROM. For playback, the company decompresses on special DVI hardware the data read in real time from the CD ROM. At the heart of the systems is a chip set developed by GE/RCA Labs which does the very fast video decompression.

The set contains two chips, VD1 and VD2. The first is a powerful pixel processing engine. The chip runs at 12.5 MIPS--a typical VAX runs at 1 MIP. It runs from on-chip RAM. It is microprogrammable, the chip's instruction set can be rapidly changed on the fly. Programs are loaded into the chip's on-board RAM. The instruction set for this microprogrammed engine is video oriented, thus allowing very rapid processing of video frames and fast graphics.

The chip is reprogrammable on the fly. It can be reprogrammed in 120 microseconds, the amount of time required for two scan lines to occur on a television screen. With this high processing power, not only can the chip decompress the image from CD ROM, it can process graphics simultaneously and display them in a window on the screen. One video image can be overlayed with another video image. The images can be moved about the screen under user control. Each of these functions can occur dynamically on the fly.

VD2 is the output display processor and is also programmable. It has multiple resolutions. It can go from 256 pixels and up to 700 pixels horizontol. It can go from 200 up to 500 pixels pixels vertically. It has multiple pixel depths, 8 16, and 24 bits per pixel. Thus, the user can create images from traditional graphics to TV-quality video.

For motion video, the 5000 bytes per frame does not provide very high resolution. The system does not provide 768 by 512 resolution in motion video unless the user opts to put the entire frame into a small portion of the screen. It provides 256 by 200 pixels of resolution for motion video. By the end of this year, the company plans to allows 256 by 240 pixel resolution. For still images, the system provides this higher resolution.

Handling Multiple Video Streams

One unique feature is the ability to handle multiple video streams. The system allows multiple video streams to be playing simultaneously. In the demo, the company showed a television screen divided into four sections and each contained a separate video image. In this application each separate stream required 1/4th the information to store 1/4th of the screen. Moreover, each quadrant of the screen can have its own audio.

The multiple channels can be dynamically mixed and controlled by application programs and user. The audio can be as high in quality as CD audio disks, but the user may want to tradeoff audio quality to store more graphics or video or vice versa. Audio is fully integrated with video and graphics.

For graphics, the system uses 15 kbytes to represent a compressed solid still frame graphic image. "However, there are many levels of uncompressed video," Mr. Kaiman says. "We can do some high quality images for anywhere from 15 kbytes all the way up to 250 kbytes for displaying a 768 by 512 pixel image.

Another feature useful in graphics applications is called video texturing. In a simulator application, for example, the user has to have the impression of being surrounded by an environment. Using a two-dimensional image, the user cannot perceive a total environment. For example, the view out the front of an aircraft and the view out the two side windows. Using video texturing, the system can create three frames in the aircraft example, one for the front window and the two others each of the two side windows. The computer can generate images to appear in each of these three frames or the system can use photographs or video of each of the three views.

The chips are so fast they can easily recalculating the entire frame from each perspective and display the image in each so that the user perceives real motion. The chips use a warp algorithm so that as the frames moves, the video for each frame gets warp to follow the movement. The chips are also capable of zoom, pan, fading, and other video manipulation feature.

Uses of DVI

Applications for the technology include training where videodisks are used today. However, DVI promises more flexibility than videodisks in these applications because the images are digital and they can be manipulated. Where this is important is in an application where a training sequence might require comparing two images. With the videodisk, the two images to be compared have to be recorded together. With DVI, each can be recorded separated and merged together. An application might be the assembly of a automobile engine. Depending on how sophisticated the application, individual parts can be recorded independently and under program control put together for the user on screen.

Other applications include using the DVI for sales presentations. Travel agencies can show travellers vacation alternatives complete with views of specific hotels and resorts. Real estates agents can show off house and neighborhoods to prospective buyers. Video games and entertainment are a natural for this technology.

Asked if the chips might be applied to communications, Mr. Kaiman pointed out that the chips are only good for the decompression of images compressed on much larger computers. Hence the chips would not serve well for two-way communications in a video-phone application. However, the decompression capability of the chips make them ideal in videotex applications where compressed frames of video could be broadcast and the user's equipment decodes the images for display. Mr. Kaiman points out that GE/RCA has not explored this avenue yet.

Mr. Kaiman had pointed out that the system runs under MS DOS and a question from the floor asked if Microsoft had any plans to support the technology by writing drivers to handle the additional video and audio capability on DVI.

Thomas Lopez, vice president of the CD ROM Division of Microsoft Corp. responded by saying that the technology was interesting and that Microsoft along with others would be evaluating it. In the interchange between Lopez and the questioner, the questioner noted that Microsoft was excluded from participating in the compact disk interactive CD-I development since Philips and Sony had defined an operating system and the hardware player as well as the format standard to be used in recording data on the disk. "The fact that DVI runs under MS DOS is something that is interesting to use," Lopez said, "but we're going to have to evaluate the technology before making a comment."

"Haven't you simply implemented your own version of the High Sierra Green Book CD-I specification in silicon," another questioner asked. "Yes, I believe we are a superset of CD-I," Mr. Kaiman responded, "but in addition, we implemented something different in that we've added motion video and that is something that cannot be done in CD-I." Mr. Kaiman added that RCA labs started the DVI development three years ago.

Another questioner asked if the chips were operational and would they be sold separately. They were designed using Silicon Compiler tools from Silicon Compilers, Inc. They were manufactured at VLSI Technology Inc. foundry in 2 micron CMOS. "The chips are working today and we currently have 18 working systems comprising three boards, one of which contains the two chips." The systems are working inside of IBM PC AT computers. "As to whether we will sell the chips, it is not clear if we will be marketing the chip sets or selling a complete system product. We are open to all possibilities at this time, but I don't have a feel for a price for the chip set," Mr. Kaiman said.

"Technically, there is no reason we cannot have the board products available by the end of the year, but the business decision will drive when we bring the product out," Mr. Kaiman says.

Mr. Kaiman also pointed out that the technology does not require a CD ROM disk to operate. It is a media-independent technology. The images could easily be stored on Winchester or floppy disk, however, the amount of storage possible on the magnetic media will be less than on the CD ROM.

"We are making a technology announcements right now," Mr. Kaiman says. "We are looking into many different business plans. There are many division of GE that are interested in this technology. We are making the annoucement now to participate in the standards process. We want to be able to join with interested parties to make this technology happen."

GE is the largest consumer electronics supplier in the U.S. It represents about 23% of the U.S. market. However, their large size has not made them eager to go it alone in developing the DVI technology.

The lab as of April 1st, will become part of SRI International. The lab was donated to SRI. GE owns the technology and GE will continue to fund the development at the labs after it becomes part of SRI. The same team of people will continue to work on the project. GE is committed to significant funding of projects at the labs on a declining scale for the next five years. For 1987, GE plans to spend $75 million in funding in 1987. Work on the video project will continue much as it has in the past.

<BACK