|
Video Processing Performance
The Fidelity of the Electronic Processing
In this section we discuss those aspects of electronic video processing that are provided specifically to enhance the incoming video signal in order to improve the displayed image. Among the more important video processing functions required for home theater displays are deinterlacing, film "3:2 pulldown", image enhancement, image scaling, and color decoding. Deinterlacing converts interlaced type signals to a progressive format, film 3:2 pulldown is a technique used to better reconstruct a film based video source, image enhancement circuits act to reduce noise and/or increase apparent sharpness, image scaling converts the source resolution to the display resolution, and the color decoding circuits translate the component input signals to the RGB signals that are necessary to drive the display. Of course there is plenty of additional processing that goes on internal to a display, but this of less interest here as it relates to less sophisticated processing or essential processing that is directly related to driving the display technology. Note that the video processing aspects of display performance discussed here don't necessarily relate directly to the display itself, but rather to the basic electronic processing of the video signals. Thus while this processing may be performed in the display, it may also be optionally implemented in a stand alone additional product positioned between the sources and the display.
Much of this processing is used to accommodate lower, or standard, definition sources, such as from NTSC television programming and DVD players, by upconverting and enhancing them for use with higher resolution displays. In some cases the upconverting improvements will be readily apparent to almost all viewers, and thus are considered very important, and in other cases the improvements will be either subtle or only apparent in specific, rarely encountered situations. We will try to give some guidance as to the relative importance below.
Unfortunately the video processing performance cannot be effectively evaluated with the WalVisions computer based test patterns, since the nature of common computer graphics cards is to present only progressive
scanned, RGB signal formats. To learn more and to evaluate your display's video processing performance with good test material, we suggest the Silicon Optix
HQV Benchmark DVD. This DVD, along with the information on the accompanying Benchmark DVD Companion document, are good sources of evaluation signals with descriptions that should nicely supplement what is more generally described below.
Deinterlacing: There are two common
formats for delivery of the video signal to a display - either by a progressive format or an interlaced format. The progressive format sends the image one horizontal line at a time, and delivers the entire image from top to bottom in one pass. This the simplest format and provides the highest image quality - as long as bandwidth (the amount of signal information that can be sent per second) is no problem. But historically bandwidth has been a problem, both for transmission and image storage (due to the program size), so the interlaced format was created. The interlaced format sends the information for a complete image at one half the data rate, while taking twice as long to send the complete image. In this case the entire image (also known as a frame) is drawn in two passes (also known as fields) from top to bottom, the first pass drawing the odd numbered lines and next pass drawing the even numbered lines. Since each pass typically takes 1/60 of a second, it will takes twice as long (1/30th second) to draw the complete interlaced image compared to drawing the same image progressively. The image retention in the eye generally "remembers" the first pass and integrates it with the next for a complete image, but whenever there is motion in the image, or when the viewer's eye moves, there will be some loss of resolution.
There are currently four common image standards for home theater in the US: the original NTSC television standard of 480 line interlaced ("480i"), 480 line progressive ("480p", which is essentially 480i deinterlaced), 720 progressive ("720p", one of the common HD formats), and 1080 interlaced ("1080i", the other common HD format). It's of note that the 720p format has about one million pixels (1280 x 720) which are refreshed 60 times per second (60 Hz), and the 1080i format has about two million pixels (1920 x 1080) which are
completely refreshed 30 times per second. Thus the rate that information is sent is approximately the same, but the 720p format loses little resolution with movement in the image, due to the faster 60 Hz refresh rate, while the 1080i format will lose some resolution with fast motion due to the slower 30 Hz refresh rate.
Thus the 720p sources may look better with motion sources, such
as sports or action films, while the 1080i sources with look
better with images that have little motion. (Of course the
ultimate would be a 1080p source at 60 Hz, combining the best of
both worlds, and while current sources don't provide this level
of both detail and motion fidelity, this can now be obtained via
the better adaptive 1080i deinterlacers.)
Essentially all current home theater displays work as progressive displays, so that the entire image is refreshed about 60
(or more) times per second, even if the incoming format is interlaced. This requires that any interlaced incoming video must be deinterlaced before being displayed - some method must be used to determine how the current image (frame) is created from the interlaced information arriving. There are simple ways to do this,
namely "bobbing" which takes the current field and doubling each line, and "weaving" which stores the first field and weaves it into the second. Bobbing is good for motion, but halves the resolution, and weaving is good for resolution, but can create a double image problems with motion. The best processors can do both - these sophisticated processors analyze the image down to the pixel level and decide what to do based on image motion - bob (or even bob and
interpolate to create new intermediate lines) for motion, and weave for static images.
Well done deinterlacing can be very good, and can make a 480i source look like 480p, and a 1080i source look like 1080p. Fortunately good 480i deinterlacers are pretty common, since DVD sources and almost all early television programming, as well as much of the current programming, are 480i and need good processing for today's home theaters, which can easily deliver that resolution. While really good 1080i deinterlacers are harder to find today, they really do make a difference for true 1080p displays, but only when viewed at closer distances where the full resolution can be appreciated. The quality of 1080i deinterlacing is a key processing feature to look for when considering 1080p displays. While in many cases the 720p and 1080i formats can appear to be pretty comparable when viewed with a variety of programming, the 1080i format processed to 1080p with a high quality deinterlacer will be clearly superior to the 720p format
when viewed on a 1920x1080 display.
3:2 Pulldown: Almost all film based material is recorded at 24 frames (images) per second, while most common video formats send frames at either 30 or 60 frames per second. Since the frame rates of film and video differ, a well defined process has been created to convert the slower film frames to the faster video frames. This process of creating video fields is simply to repeat the film frames is a sequence that goes 2x-3x-2x-3x-2x-3x, etc., thus this first film frame is used for the first two video fields, the second film frame is used for the next three video fields, etc. If a video deinterlacer can recognize this sequence, it can then precisely reconstruct the film frames in a process known as "3:2 pulldown", and this eliminates the need for the motion detection processes that can sometimes degrade the image slightly.
It is of note that films may be edited when they are converted to video/DVD, so there may be edited sections in the movie that deviate from the 3:2 cadence. The best of the processors will quickly adapt to these changes and provide seamless transitions, while average processors may show a glitch at the edited transitions. Also note that some lower cost DVD players with progressive outputs have poorer 3:2 pulldown/deinterlacing circuits than many display devices. Thus we recommend that you don't automatically put the DVD player in the progressive mode - first
compare the DVD 480i output (using your display's deinterlacer)
to the DVD progressive output (480p, 720p or 1080p). Note
that the HQV Benchmark DVD (noted above) is excellent for this
type of comparison. Simply chose which DVD output format
that gives the best images, particularly with motion.
Image Enhancement: There are a number of processes that fall into the image enhancement category, the most notable being noise reduction and edge enhancement, which is sometimes referred to as a detail or sharpness control. This is an area of processing that can really clean up some marginal sources, but both removing random noise and by enhancing the apparent image resolution. But be careful, sometimes parts of the image can be mistaken for noise, and the added sharpness can create unnatural detail and "ringing" (or "ghosts") at the edges of objects.
Image Scaling: Image scaling is necessary whenever the image must fill a fixed resolution display, such as DLP, Plasma, LCD or LCoS, and the input signal format doesn't match the display resolution. An example is when a 480p signal (720 x 480 pixels) from a progressive DVD player is displayed on a native 720p (1280 x 720 pixels) projector. In this case, a image detail that might involve about 6 pixels at the incoming 480p resolution will have to be remapped to about 10 pixels on the 720p display. Poor scaling might simply "enlarge" the original image, rather than finding the best fit to the increased number of pixels, or might create a "blocky" effect when there is motion in the image. Fortunately most scalers today are pretty good
in this regard, so in most cases scaling won't be a problem.
If a display's scaling is turned off, by selecting a "true" or "native" format in the control menus, the image will map pixel for pixel, and will only
just completely fill the display when the source and display resolutions exactly match. It's also interesting to note that sometimes when the source and display resolutions match, there still may be scaling involved! This can be due to "overscan", in which the image is enlarged a little so that any irregularities at the image edges are not seen. For instance, a 720p (1280 x 720 pixels) signal may by enlarged/scaled to about 1330 x 750 pixels to achieve 2% overscan per edge, and then only the center 1280 x 720 pixels of the enlarged image shown on a 720p display.
Color Decoding (SDTV
vs. HDTV): Modern video systems are made to be as efficient as possible to minimize both the bandwidth necessary for transmission as well as the video storage/memory necessary for archiving. One of the ways to be efficient is to send a full bandwidth luminance (Y - black and white) part of the signal, along with lower bandwidth "color difference" (PB, PR) signals. This is more efficient than sending three full bandwidth signals (red, green and blue), and just as effective since
while the eye needs lots of luminance detail, it can be
satisfied with lower color detail information.
This does
require a little extra processing, however, both at the point of
signal origination and also at the display, but the small amount
of additional circuitry isn't a problem, and the bandwidth and
storage savings are well worth it. Except now add in one
other small detail - the SDTV (standard definition, such as DVD)
and HDTV encoding standards are not exactly the same! Now
if a display is only capable of HDTV decoding, like many are,
when a DVD is displayed the color shades will be somewhat incorrect! See the "color bar" figure below - the center row of bars is the correct reference and what you would get if you encode-decode with the same standard, while the top and bottom rows are the resulting colors from "cross-decoding". While the differences aren't dramatic, and most people don't
readily notice, they are easy to see and noteworthy.
Thus if you're interested in the highest color accuracy from your display, make sure that it can automatically detect the signal type and provide the correct decoding.
|
These color bars have been
HDTV encoded, but SDTV decoded. |
 |
|
These
color bars are the source and thus correct colors, and are what you will get with matching encoding and decoding processes. |
|
These color bars were
SDTV encoded, but HDTV
decoded. |
|
