DVDO's PureProgressive technology converts video streams from standard interlaced video to progressively scanned video. The following article describes why progressive scanning is inherently superior and how DVDO's exclusive PureProgressive performs this conversion with maximum video quality.
Interlace vs. Progressive Scanning
Interlace scanning is used in today's standard analog televisions. An interlaced TV "paints" the lines of a frame in two separate passes. Half of the lines are drawn in the first pass (the even lines), and the other half (the odd lines) are drawn in the second pass. First devised so that early TVs could have decent resolution with the limited transmission technologies available at the time, interlaced scanning has several unfortunate side effects which are discussed below. Progressive scanning paints all of the lines of a frame in one top to bottom pass. This is used where transmission bandwidth is not an issue and where the highest quality image is required. None of the interlaced side effects are present with progressive scanning.
The first major problem with interlaced scanning is that the image may visibly flicker if the screen is large enough that it represents a significant portion of a person's viewing angle. Even with small screens, sharp edges on objects may flicker. In addition to the flicker, vertically adjacent horizontal lines are not from the same field so motion occurring over time causes a spatial displacement on the display.
This motion displacement over time is usually not noticeable to most television viewers because the time displacement of the fields causes the "older" field to fade in intensity. However, on high resolution displays or on devices such as LCDs or plasma panels which do not fade, an interlaced image will contain noticeable motion artifacts. Just as critically, this displacement means that, during motion, the screen is only able to resolve half as many lines as it can for still images. For a standard TV, that means that the theoretical vertical resolution of approximately 480 lines actually translates to more like 240 lines during movement. For a medium that is designed for movement, this is a severe limitation.
Deinterlacing is the process by which interlaced video is converted to progressively scanned video. Devices for performing deinterlacing are available for tens of dollars for low quality techniques or for many thousands of dollars for very sophisticated techniques. The low cost techniques are frequently used in progressively scanned TVs or projectors. High quality algorithms are used typically in Line Doublers designed for high-end home theater markets and generate very high quality video.
inexpensive deinterlacers simply put fields together, creating an output frame containing
even lines from one point in time and odd lines from 1/60 second later. Any motion between
these two fields will result in the motion artifacts illustrated above.
To avoid these artifacts, some deinterlacers simply scale each of the fields up to the entire frame size, interpolating between the existing lines. Unfortunately, this also significantly reduces the vertical resolution of the image, resulting in softening of the picture with loss of image detail.
One method of avoiding this softening when it isn't needed is to determine if there is any movement between fields by comparing each of the fields with its counterpart in a previous frame. Further refinement of this algorithm would be to apply the softening filter only to portions of the image that are in movement. This is referred to as "motion adaptive" deinterlacing.
The most advanced and best quality Line Doublers are designed to also take advantage of the "2:3 pulldown" technique that is used to transfer film to video. During this transfer, the first film frame is captured onto 2 video fields (first even, then odd lines are scanned) then the second film frame is captured onto 3 video fields (even, odd, even). As this is repeated, you can see that two 24Fps film frames (for a total of 1/12 of a second) are captured onto five 60fps video frames (for a total of 1/12 of a second). A deinterlacer can examine a series of frames to detect this sequence and thereby determine that the original, pre-video source of this sequence was film. It can then reassemble the original progressive frames from the partial interlaced frames with no loss of resolution or with no introduction of motion artifacts.
DVDO's PureProgressive technology performs even more advanced techniques than those described above. Performing over six billion arithmetic operations per second on the incoming video stream, PureProgressive uses the data from four video fields to determine not only which portions of the image are in motion, but also what type of movement exists in each (using our proprietary modified Fourier technique), and how best to generate a progressive image with maximum picture detail for that portion.
PureProgressive also performs excellent 2:3 pulldown detection. In addition, it also recognizes the "2:2 pulldown sequence" used for converting computer graphics to video. For film and computer graphics sources, PureProgressive flawlessly reassembles the original progressive frames. Unlike some other processors that advertise similar capabilities, PureProgressive reliably detects the source type even with noisy signals and poor video sources.
One problem with 2:3 pulldown detection is that if the 2:3 sequence changes it will be at least several frames before most processors will detect the change. In the meantime though, the processor has been blindly and incorrectly assembling frames, resulting in severe motion artifacts in the output. PureProgressive avoids this by immediately detecting when the source type changes preventing it from ever generating these artifacts.
PureProgressive Processing Architecture
The question that remains to be answered is how is it possible for DVDO to match the performance of Line Doublers costing many thousands of dollars in a package that costs only a fraction of that. Compared to these high end processors, PureProgressive performs comparable or, in many cases, superior deinterlacing and certainly produces output that rivals or exceeds these boxes.
There are three primary reasons why this level of performance is possible.
The first reason is that, from the start, it was our goal to attain the very high quality of the high-end home theater line doublers but at a low price. We felt that the time was ripe for a line doubler that was every bit as good as those doublers but with a price that most large screen TV owners could afford. In addition, driving the cost of this technology that low would allow PureProgressive to be embedded in TVs, DVD players, A/V receivers and satellite receivers thereby eliminating the requirement for a standalone external line doubler in many systems. To accomplish this meant that, in addition to designing the best deinterlacing algorithms, we had to design our own chips from scratch to implement those algorithms. Using a slew of expensive chips from other manufacturers was not an option.
Second, though the processing algorithms themselves were designed to deliver very high performance, they were also designed from the start with implementation in mind. In other words, the algorithms were not designed in isolation - our video scientists are also our chip designers and they have many years of experience doing both.
The third reason is the architecture of the chip itself. One of the challenges encountered with processing video is that two dimensional horizontal and vertical processing is required, but video data is presented to the processor as a one dimensional stream. A solution to the problem of two dimensional processing is to embed memories on chip that store a number of horizontal lines. These memories are then read in parallel, producing a set of vertically aligned video samples for processing. The quality of the processing algorithms are frequently dependent on how many horizontal lines of data are available. Quality relates directly to the number of line memories on chip.
Line memories are typically large structures that have a significant impact of chip size, power consumption, and cost. DVDO has developed a unique processing architecture, for which patents have been applied, that offloads the on-chip memory requirement into an external low cost commodity SDRAM. The high bandwidth requirement in and out of line memory is solved by having much smaller memory structures on chip while offloading most of the storage to the external SDRAM.
The result of these innovations is a reduction in the on-chip line memory requirement by about 90%; this memory savings translates directly into reduced chip cost and reduced chip cost ultimately translates into reduced overall system cost.
DVDO, Inc., is a privately-held company that designs, manufactures, and markets chip and system-level solutions for the next generation of Digital Television and Digital Video electronic products. The target market spaces for DVDO's products include Digital Video Disc (DVD), Digital Television (DTV), Digital Satellite System (DSS), Home Theater, Automotive, and Airline Entertainment applications. DVDO is the creator of the proprietary PureProgressive Image Enhancement Engine Digital Video Technology, a new high-performance and cost-effective solution for chip-level digital video processing and enhancement. DVDO has a total of fifteen U.S. and international patents pending on this new technology.
Copyright © 1998 - DVDO, Inc. - 1129 Dell Avenue - Campbell, California 95008 408.364.3836