Line Doublers, Progressive Scan
By painting the scan lines on the picture tube faster the picture is made smoother and flicker is reduced. But getting the maximum picture quality from interlaced video takes more than just painting each scan line on the picture tube twice.
January 2009. We are rebuilding our web site. Please bear with us regarding broken links.
Last update 2/06
In a Nutshell
Interlaced scanning, painting 480 scan lines on the screen as two 240 line fields every 30/th of a second (for NTSC), was invented to get the best combination of horizontal resolution, vertical resolution, and freedom from flicker using the limited bandwidth of the TV broadcast channel.
Spurred on by the personal computer industry, which found the flicker of interlaced video to be very objectionable with fine detail, the TV industry adopted 480 line 60 frame per second progressive scan as the next step up in quality for a smoother picture and moving subjects looking more solid. We have the same 525 scan lines per frame with 480 holding the picture, but we get the whole picture every 60'th of a second as opposed to just the even lines or just the odd lines.
The unit needed to take a standard (interlaced) broadcast and create a progressive scan video signal is generically called a line doubler since the number of scan lines delivered to the TV every second is twice what we used to have. The simplest line doublers merely deliver each scan line to the TV twice; those that do more to yield a better picture are specifically referred to as de-interlacers or progressive scan converters.
Other types of doublers double the scan lines yet again (to 960 for NTSC) so the scan lines are less obtrusive on very large screens.
The TV set or monitor must specifically accept the progressive scan video signal, most older TV sets do not. Almost all U.S. HDTV sets accept 480p progressive scan.
Converting interlaced to progressive scan requires different processing processing compared with line doubling just to increase the number of scan lines per frame for large screens. To convert to progressive scan and preserve both detail and fullness, material from two consecutive fields should be combined for "stationary" subject matter. For "moving" subject matter a needed portion of an even line should be a blend of the odd lines immediately before and after, and vice versa.
Progressive scan by itself does not improve resolution. We still have the same 480 (for NTSC) illuminated scan lines' worth of picture detail.
Flicker reduction is achieved by painting (drawing) all the scan lines more often, not by progressive scan in itself.
Mediocre de-interlacers and line doublers detract from either or both horizontal and vertical resolution.
We don't have specific brand names and models that use or don't use particular methods described here. You will have to evaluate the TV picture yourself.
Why Interlaced Scan?How Does It Do It? (in order of sophistication)
Plain line doubling
Bob
Weave
Motion Adaptiveness
Motion CompensationSpecialized methods:
Diagonal Processing
3-2 Pulldown Sensing & Optimizing
DefactoQuality differences
Pixel Width
Granularity, Macroblocks
Motion Artifacts
The Issues Addressed
1. In ordinary interlaced video the odd lines start to fade while the even lines are drawn and vice versa. For "stationary" subjects some people notice the scan line structure of the picture because alternating scan lines are of different brightnesses and/or seem to pulsate.
2. In ordinary interlaced video, for moving subjects half the picture (say the odd scan lines) "moves", then the other half (the even lines) leapfrogs over it, and then the first half does a leapfrog and so on. Especially if a subject moves over a contrasting colored background the scan line structure is visible as the partly faded background scan lines show up between the the subject scan lines.
3. Especially on larger screens, the scan lines may be thin enough that dark gaps are visible between them.
If the video signal is interlaced to begin with, the "first doubling" converts the interlaced scanning to progressive scanning only. It addresses issues #1 and #2, but not #3 above. The line count, as in 480i, stays the same; in concise technical terms we convert 480i (NTSC) to 480p. Tripling, or a second doubling, would be needed to address #3.
If the video signal is progressive to begin with, line doubling addresses issue #3. Here we would say we convert, say, 480p to 960p.
Different mechanisms (formulas, algorithms) are needed to convert to progressive scan versus simply increase the number of visible scan lines.
Before buying a progressive scan converter (de-interlacer) or line doubler, shop around to view the picture quality. Also, be aware that it will not work with a TV set not specifically intended for progressive scan, called 480p in the U.S. Currently all "HDTV ready" sets accept progressive scan from a de-interlacer.
Why Interlaced?
It was a tradeoff between number of scan lines for vertical resolution, fineness of details within a scan line for horizontal resolution, and number of frames per second for minimum flicker.
In the early days of TV, there was a severe limit to how many dark to light transitions (read: pixels across a scan line) could be transmitted each millisecond* compared with today. In order that moving objects not look smeared, the phosphors on the picture tube had to fade.between visits by the electron beam. At the chosen resolution (for NTSC) of about 480 scan lines each of which could be divided into as many as 440 parts, the entire picture could be redrawn (repainted; refreshed) 30 times a second. At this rate, the top of the picture began to fade before the bottom was completely drawn. Then someone came up with the idea of drawing the odd scan lines first, then drawing the even scan lines, then the odd scan lines, etc, which lmost eliminated that rolling fading flicker,.and obtained a much better looking picture for the same amount of transmitted information.
Some viewers still noticed flickering of extremely small details and very thin horizontal lines. But since most of the picture consisted of not so fine details, this was not considered too objectionable. Much research has gone into chemically designing the phosphors to fade at an optimum rate. From the early days of TV to today, varying amounts of bleeding picture content onto the next scan line (called vertical filtering) which by definition reduced vertical resolution and softened the picture was done. Usually vertical filtering was done in the camera, where the more of it was desired, the wider the horizontal strip on the "film" was taken into account and averaged out to create each scan line.
Some of today's HDTV is interlaced and for the same reasons, limits on the amount of picture detail that can be transmitted in the allotted broadcast channel, and limits on the electronics imposed by costs.
*There was no memory in those days. The "pixels" had to be put on the picture tube the instant they arrived; putting fewer of them on one scan line (no fine subject detail there) did not mean that more could be put on the next scan line.
Why Progressive Scan?
The primary intent of progressive scan is to refresh the screen more often.
Up until the late 1980's, flicker on computer screens was very noticeable since single scan line details made up a much larger portion of screen content. Also with memory as a limiting factor, consumer PC's only had about 240 scan lines of picture information which incidentally hid most of the flicker. The "regular VGA" standard was based on NTSC, exactly twice the scan rate using the same 525 scan lines per frame and progressive scan, with 480 scan lines holding the picture and with up to 640 details on a scan line..
As larger TV screens were developed, more viewers started noticing the flicker due to the fading phosphors when the electron beam visited any given spot on the screen only once every 1/30'th of a second. When you "see the scan lines" you are really seeing the even gaps between the odd scan lines or vice versa, as the phosphors fade between refreshes. On small screens in the early days of TV, the electron beam was thicker than 1/480'th the screen height so these gaps were not as noticeable.
What kind of TV you need?
The standard TV of the 20'th century cannot do progressive scan. Almost all computer monitors made since 1990 can. For the computer monitor you do need a "box" connected between your VCR (or DVD player) and the computer monitor to convert the interlaced video to progressive scan video and also adjust the color breakdown to match.
All "HDTV ready" sets I know of will accept progressive scan standard definition (SDTV) video inputs. One problem is that some models only display progressive scan pictures in a 16:9 shape, and I suggest not buying any of those models. You don't want to be forced to watch traditional 4:3 shaped pictures unnaturally stretched out.
Plain Line Doubling
The simplest line doubler does produce a progressive scan video output. It sends each source scan line's content to the TV twice. So for the first video frame, the source scan line content goes like 1, 1, 3, 3, 5, 5, and so on. The next frame has source scan line content 2, 2, 4, 4, 6, 6, and so on. Painted on the screen, the source scan line 1 content is commingled with the scan line 2 content, and so on. As a slight refinement, the even frame content should be staggered downward by one scan line so the painting on the screen is as follows:
Position 1 -- Content of (source) scan line 1, or black alternated.
Position 2 -- Content of scan line 1 or scan line 2 alternated
Position 3 -- Content of scan line 3 or scan line 2 alternated
Position 4 -- Content of scan line 3 or scan line 4 alternated
Position 5 -- Content of scan line 5 or scan line 4 alternated
We estimated we lost 1/3 of the vertical resolution using this method. We could resolve a test pattern with every third line white with some softening. (A test pattern with every other line white was a total blur.) This method does not require holding an entire field in memory. Product example: Instant TV by AIMS Lab (discontinued).
This is similar to plain line doubling except that instead of outputting a scan line for the second time as-is, the scan line is electronically mixed with the next scan line. So for the first video frame we have content as follows: line 1, lines 1 and 3 mixed, line 3, lines 3 and 5 mixed, and so on. The mixing, or commingling, process is referred to as interpolation since each video field is taken line by line and, say, given scan lines 1 and 3, we guess what scan line 2 should look like. The process is also called "bob" because to some viewers, the entire picture seems to vibrate up and down slightly. Still, the fact we do commingling means that the finished picture is softened compared to a system where no commingling is done. The overall process is also called scaling since we are starting with a field or frame with so many scan lines and ending up with a field or frame with a different number of scan lines.
Two Field Bob With Stagger; One Field Bob
Properly done, the bob method has the even frames staggered (offset) down by one scan line This is needed so that the original even scan lines land in the even positions on the screen. When both the odd and even fields are displayed the result is as follows:
Position 1 -- Content of (source) scan line 1, or black, alternated
Position 2 -- Mixture of lines 1 and 3, or content of line 2, alternated
Position 3 -- Content of line 3, or mixture of lines 2 and 4, alternated
Position 4 -- Mixture of lines 3 and 5, or content of line 4, alternated
Position 5 -- Content of line 5, or mixture of lines 4 and 6, alternated
The stagger is also called a "half line phase shift" since on a CRT the topmost line for an even field drawn on the screen is a half line starting at the screen center.
Without staggering, the original scan lines always land in the odd positions and the interpolated scan lines always land in the even positions. This causes the vertical resolution to be no more than that of a single field, 240 lines for NTSC or 540 lines for 1080i HDTV. It is as if 1080i were treated as 540p.
We have also heard of systems that ignore (discard) the even fields, scaling up and outputting each of the odd fields twice (one field bob) This also (and by definition) loses half of the picture information, yielding vertical resolution no more than that of a single field, and also slightly jerkier motion.
Each video frame is constructed using the odd scan lines from one field and the even scan lines from the next field (or vice versa). After all, interlaced video is referred to as having so many scan lines per frame, half of the scan lines (the odd lines) painted first and the other half painted next. Overall the picture appears sharper. There is no commingling of scan line content, so there is no loss of vertical resolution.
A new problem arises with the weave techinque. Subject motion may have occurred in the 1/60'th second between consecutive video fields. Because progressive scan has eliminated the flicker and fade of the odd scan lines as the even lines are drawn, feathered side edges as the subject moves are more noticeable. We are seeing the subject in one position corresponding to the odd scan lines and in a second position (like a double exposure on film) corresponding to the even scan lines.
3-2 pulldown sensing and optimizing (discussed later) is a specialized form of "weave" taking advantage of the fact that matched even and even fields predominate in film source video.
If the subject is moving, we really have no excellent choice, either accept a softer picture and interpolate intervening scan lines using only the material in one field, or accept the feathered edges (combing) from weaving in scan lines from the next field that might not match. Fortunately when the subject is moving, a slight loss of sharpness is less noticeable compared with a stationary subject. Meanshile for stationary subjects, weaving is the better choice for maximum sharpness, or resolution. So the best technique with reasonable cost to date is to use both bob and weave, bob for the "moving" parts of the picture and weave for the "stationary" parts.
What sets aside an excellent de-interlacer from a mediocre one is being able to guess what parts of the picture represent moving subjects. (The constant changing between methods of generating the intervening lines is a process generically referred to as adaptiveness. The term "motion adaptive" is used here because the choice of strategy happens to depend on whether the subject material is stationary or in motion.) To better recognize when a "different" line is different due to subject motion or different due to fine subject detail or "different" due to video noise (snow) requires having three or more fields' worth of video information on hand and continuously sizing up several scan lines before and after the one being processed..
The best de-interlacers can switch between bob and weave many dozens of times during each scan line.
Unfortunately deciding whether subject matter is stationary or moving is never perfect.
Quality Gradation
The best methods of motion adaptiveness literally go through the video frame pixel by pixel when deciding when to "bob" and when to weave. The best methods can weave in material from either the following field or the preceding field.
The next best method also does motion adaptiveness pixel by pixel but only uses the preceding field for the parts of the frame that need material woven in. It would take 1/60'th second more for a cessation of motion to be detected, when what was the "current" field becomes the previous field and there is a new current field.
Vastly inferior methods that barely fit the definition of "motion adaptive" persist to this day. Some inferior were introduced in an effort to compete with already existing superior methods. Quite common was a situation where, if enough of the subject matter was determined to be stationary, the entire field was woven with its predecessor, otherwise the entire field was bobbed. The result was, when something started moving in an otherwise stationary scene, the entire scene became less sharp. Sometimes the frame was divided into rectangular zones where, within any given zone, if everything was stationary then subject matter was taken from the previous field to be woven. The keen eye will still see checkerboard squares change from soft to sharp as motion stops or starts within each block. The finer divisions the picture is divided into, or better, dealing with variable sized fragments of individual scan lines with no checkerboard or zone pattern, the better the picture. We are told that a major difference between the Silican Image iScan (all models) and the much less expensive Viewsonic VB50, is that the iScans have a much finer granularity, so to speak. For evaluation, observe diagonal boundaries, both in live pictures and cartoons.
What seems to be a new method, not the best, was recently discovered. The de-interlacer began by weaving two consecutive fields. Then using the methods used for determining moving parts of the scene, it identified and blended the parts of the image where motion was represented so as to hide combing. Motion blur was more pronounced compared with bobbing the parts of the image representing motion, and objects appeared to be slightly larger when in motion. Some improvement was had by interpolating only even lines using only material from odd lines when an odd field was being processed, and interpolating only odd lines using only material from even lines when an even field was being processed.
A TV set or stand alone de-interlacing unit can use motion adaptiveness in a built in Y/C separator (comb filter), in de-interlacing, in both, or in neither. The purchaser should confirm this prior to purchase of equipment.
A relatively new and expensive technique to use weave more often and bob less often is called motion compensation. The de-interlacer analyzes picture content to guess where in the next field an object has moved to. It constructs a full video frame by taking the next field, "moving the subject back" to where it was in the current field and then weaving in the intervening scan lines from the next field. Large objects moving horizontally are the easiest to guess the position of. Patterns that repeat from left to right are the most difficult and the de-interlacer should fall back on the other techniques we have described here if it "is not sure" of where the subject has moved to.
Diagonal Processing
Diagonal lines and edges can be made crisper using additional processing compared with straight interpolation to generate intervening scan lines. Probably the best known example is Faroudja's DCDI processing. Fundamentally this is accomplished by analyzing content possibly several pixels removed from the location being interpolated, as opposed to simply taking an average of the scan line above and the scan line below
3-2 Pulldown and Inverse Telecine
U.S. movies are typically shot at 24 film frames per second. In order to commit the film to NTSC video (approx. 60 fields per second), a repeat pattern, called 3-2 pulldown (or 2-3 pulldown) is used:
Video field 1 -- Film frame 1 odd scan lines
Video field 2 -- Film frame 1 even scan lines
Video field 3 -- Film frame 1 odd scan lines again
Video field 4 -- Film frame 2 even scan lines
Video field 5 -- Film frame 2 odd scan lines
More on 3-2 pulldown, click here.
When the de-interlacer has good 3-2 pulldown sensing and optimizing (recognition and retention) it can almost guarantee obtaining matching odd and even fields for each progressive scan video frame and.use the weave method for maximum vertical resolution. For any given field, either the field following or the the field preceding (or both) matches in content when 3-2 pulldown is present.
One problem arises if the 3-2 pulldown cadence in the source has irregularities. The de-interlacer should switch to motion adaptive methods but sometimes has only more primitive methods available to fall back on. There are some inferior methods that have been known to continue on indefinitely using the cadence that was initially derived while the actual cadence was different. Weaving of mismatched fields then occurs.
Although 3-2 sensing and optimizing is sometimes called inverse telecine, the video as output is still (approx.) 60 fps, not 24 fps. and the 3-2 pulldown cadence is still present.
All DVD players initially decode the video program as interlaced. Progressive scan players have a built in de-interlacer. Most DVD programs of 24 fps film contain flags that tell a player's de-interlacer where matching odd and even frames are positioned. Unfortunately the flags for this purpose are not always correct. The ideal de-interlacer for a DVD player senses flags in the video material for film source but also spot-checks the video content (using pixel analysis) to verify that the flags are correctly positioned, and has motion adaptive de-interlacing as yet another stage to fall back on.
An unrelated use for actual inverse telecine methods is producing DVD's where the only available source was video with 3-2 pulldown. For NTSC, interlaced video of 24 fps film has four of every five fields unique, for progressive scan, two out of every five frames are unique. The inverse telecine process is used to discard the redundant fields/frames yielding a 24 frame per second (or 48 field per second) video stream. This is done so more material can fit on the disk giving greater playing time. (The player regenerates the 3-2 pulldown as described earlier.)
When the de-interlacer uses both the following field and the preceding field for motion adaptiveness, the result can come close to 3-2 pulldown sensing and optimizing.
(A telecine machine makes television from cinema, that is, takes movie film and generates a video signal.)
Sample Methods
Method 1 (starting with a threesome and a pair follows)Take field 1, determine it matches field 2, weave them together to become frame 1Take field 2, determine it matches field 3, weave them together to become frame 2
Take field 3 determine it does not match field 4 but that it does match field 2, weave with field 2 to become frame 3
Take field 4, determine it matches field 5, weave them together to become frame 4.
Take field 5, determine it does not match frame 6 but that it does match frame 4, weave with field 4 to become frame 5.
In Method 1 the de-interlacer may keep a running count of a 3-2-3-2 cadence and do just spot checks for the above sameness determinations, changing to video mode de-interlacing should an unexpected field mismatch be encountered..
Method 2 (also starting with a threesome and a pair follows)Knowing that a 3-2-3-2 cadence is in progress and that we we are starting with a threesome, take field 1, weave it with field 2, and use the result for frames 1, 2, and 3.Still knowing that a 3-2-3-2 cadence in progress and that we are up to field 4 which is first of a pair, weave it with the next field (#5) and use the result for both frames 4 and 5.
In Method 2 the de-interlacer would do additional spot checks to detect loss of cadence showing up as unexpected field mismatch, and change to video mode de-interlacing if needed.
Method 2 more resembles an actual inverse telecine process although equally good results can be obtained whether Method 1 or Method 2 is being used.
Mixed Film and Live Video
In recent years, material using both 24 fps film source and 60 fps video source combined in the same frames has been produced. This might be part of special effects in movies, or encountered where animated material and live action were combined.
Up until recently, 3-2 pulldown sensing and optimizing was applied only to the entire video frame if at all. If the de-interlacer uses its film mode, embedded live video parts would exhibit combing. In 2006 some processors have been introduced that perform motion adaptive de-interlacing for some parts of the frame and 3-2 pulldown sensing and optimizing for other parts of the frame on a pixel by pixel basis to better handle mixed film and live video content.
Current State Of The Art
The performance plateau for "good" de-interlacing has been relatively unchanged since 1999 when the then DVDO Inc. introduced its iScan Plus which could be purchased for an affordable $600. The plateau consists of 3-2 pulldown sensing and optimizing for the whole frame falling back on pixel by pixel motion adaptiveness for the whole frame for film source, and pixel by pixel motion adaptiveness for live video source, selected automatically. Today (early 2006) this plateau has been met for both standard definition and high definition formats.
We would estimate that a third of the SDTV products and less than 5% of HDTV products needing de-interlacing meet this plateau with some minor quality variations for such things as diagonal enhancement and occasional misjudgments in applying bob versus weave..
Motion compensation is still in an "experimental" stage. Currently no consumer grade products use it.
As of this date (2006) only a few motion adaptive de-interlacing solutions have been introduced for HDTV. De-interlacing is required for 1080i to 720p as well as for 1080i to 1080p. On the other hand there are even a few instances where the even fields were ignored (discarded) and the odd fields used twice for "bob" de-interlacing. The concept of motion adaptiveness is the same for HDTV as it is for SDTV; it is only that a much larger amount of material needs to be processed.
Triplers, Quadruplers, Scalers
There are also line triplers that triple the number of scan lines, and so on. These were originally used for commercial applications such as for video in theaters where picture size of ten or more feet was desired. Broadly speaking, the purpose of a line tripler is to fill the gaps between scan lines and make a video picture look as if it was not constructed as scan lines while not making the electron beam so fat as to reduce horizontal resolution.
Most line triplers, line quadruplers, and similar vintage line doublers did not use any sophisticated de-interlacing techniques. Considerable softening of the picture was seen because each video field was treated independently and scan line content blended optically, for example (for a tripler):
Position 1 -- Scan line 1 or black
Position 2 -- Scan line 1 or scan line 2
Position 3 -- Scan line 1 or scan line 2
Position 4 -- Scan line 3 or scan line 2
Position 5 -- Scan line 3 or scan line 4
Position 6 -- Scan line 3 or scan line 4
Position 7 -- Scan line 5 or scan line 4
Once we get a video signal we can regard as progressive scan, doubling the lines (again) is easy. We can use the plain doubling or bob methods exclusively from this point on. All of the vertical resolution is present and synthesizing the added lines does not lose resolution.
Some line "doublers", more correctly called "scalers", allow the user to select doubling, tripling, etc. based on the size of the gaps between scan lines. Selecting much too high a multiplier softens the picture vertically (loses vertical resolution) because scan lines overlap. Even fractional upward scaling can be done. This involves a complicated blending process also known as upconversion (usually better) or repeating some lines but not others (less expensive). The advantage of fractional scaling is to get exactly the number of scan lines needed to avoid both gaps and overlaps given a fixed electron beam spot size or to match a fixed number of rows of pixels in an LCD, DLP, or plasma panel.
Actually a 480p signal can be treated as a 960i signal by the TV. Some TV sets inadvertently do this because they have a small vertical jitter causing each field of scan lines to be slightly staggered. This hides gaps between scan lines at the expense of having some interlace flicker come back..
Line doublers, triplers, scalers, etc. cannot be used with "any old" TV set. The TV set, monitor, or video projector must be capable of matching the scan rate and the scan method (interlace vs. progressive) that the line doubler outputs.
De-Facto De-Interlacing
LCD screens do not fade like phosphors on a picture tube. If interlaced video is fed to an LCD system that accepts it, the odd scan lines (rows of pixels) remain lit at full strength as the even scan lines are filled in and vice versa. This produces, with no processing, the weave method of de-interlacing. This method is not common since the LCD system usually does not have the separate circuits needed to address the rows of pixels alternately for interlaced scan.versus consecutively for progressive scan.
Motion Artifacts
When progressive scan video is constructed using the weave method exclusively, viewers may see the feathered side edges of moving subjects, as the even scan lines show the subject in a dfferent position from the odd scan lines. This effect is an example of what is referred to as a motion artifact. This effect is also called combing (no relation to comb filter) because the feathered or serrated edges suggest the teeth of a comb.
Frame Doubler vs. Line Doubler
The term "frame doubler" is sometimes used to describe good de-interlacing units because a complete frames are developed where half a frame (one field) was before, and then the output is delivered to the TV. The units are also still called line doublers because twice as many scan lines come out as are put in.
Frame Retention
Frame retention refers to a buffer, or memory, to hold one or more fields or frames while new arriving video information is added to it.
Incoming video is a stream of information. Without memory, the conversion circuits would have to take each line's worth of information as it arrives and use it, or lose it forever. All de-interlacers and with any degree of sophistication have frame retention using memory (similar to computer memory) to hold on a rolling basis at least one field (half picture) of information. Then the "real" intervening line from the previous field (say line 2) will always be on hand if the logic decided not to blend the lines above and below (say lines 1 and 3) to synthesize that line. Some de-interlacers hold more than two fields in memory. This allows their logic to better detect motion and also allows an intervening line to be taken from either the preceding or the following field if the subject just started moving or just stopped moving, respectively.
Vertical Resolution Loss
For interlaced video, moving subject matter has only half the vertical resolution because each snapshot of the subject has half the scan lines, even or odd. Here, blending scan lines or painting them twice is a good educated guess as to what the subject should look like and the resolution stays the same. The lesser vertical resolution is less noticeable especially considering that during the original filming, blur due to a relatively slow camera shutter speed does exist. But for stationary subjects, there should not be any adjacent line blending.
Simple de-interlacers that don't remember an entire field (are not field retentive) must rely on the bob method (blending scan lines electronically) or repainting them twice (blending them optically) all the time. They suffer from vertical softening (loss of vertical resolution) over the entire picture. Side by side, still subject matter shown in the original interlaced video will be clearer than the progressive scan converted result using the same blending throughout the video frame.
Horizontal Resolution Loss; Pixel Width
All line doublers, de-interlacers, and scalers have a "pixel width"; they use digital processing and must break up each video scan line into pixels. An insufficient pixel width results in loss of horizontal resolution. At 1.5 times the maximum number of pixels across for the source material, there is no noticeable loss in horizontal resolution. We don't have pixel width information and don't know how best to obtain it, but we suspect that one of the reasons for mediocre line doubling is a relatively low pixel width. We consider 512 and under to be insufficient for DVD and marginal for TV shows, 640 to be marginal overall, 720 to be reasonable, and 858+ to be superior.
Progressive scan requires better electronics and cables, specifically with the ability to handle frequencies twice as high. With twice as many scan lines being drawn every second, each scan line with all its light to dark to light changes must be completed in half the time compared with the original interlaced video. Horizontal resolution is directly dependent on the ability of the electron beam to change from light to dark to light quickly. Deficiencies in electronics and cables will show up first as lower horizontal resolution..
Analog to Digital Conversion
Because all doublers, de-interlacers, and scalers perform their work digitally, any incoming analog video signal must be converted to digital. If the next video device expects an analog input, the video after de-interlacing and/or scaling needs to be converted back to analog.
A little known fact about digital video is that repeated converting back and forth between analog and digital degrades the horizontal resolution each time. Stand alone de-interlacers and line doublers are at a disadvantage when they use conventional component video and S-video cabling, which are analog signal paths. Analog video sources such as broadcasts and laser disk are less susceptible to this form of resolution loss provided that the source material was not digital at any earlier time (such as being a "D1 master") in its life.
As of 2004, many models of TV sets, stand alone de-interlacers, and DVD players offer digital progressive scan connections. When the digital progressive scan output of a DVD player is connected to the digital input of a TV, there is no analog to digital conversion occurring anywhere in the video signal path.
But digital inputs and outputs need to be properly matched. There are currently three common digital video formats: DVI, HDMI, and Firewire (IEEE 1394). Within these formats there are individual video "resolutions" such as (one version of 480p) 640 x 480 x 60 frames per second non-interlaced or (1080i HDTV) 1920 x 1080 x 30 fps interlaced. Also, some viewers may feel that DVD's 720 x 480 native format output as analog gives a better picture than a 640 x 480 digital output.
When a digital video signal is converted back to analog, a "footprint" of the horizontal pixel structure and spacing remains. Converting again to digital, if the new pixels split (straddle) rather than line up with the old pixel structure, this is a major cause of softening, or loss of horizontal resolution. See, also, our discussion of Kell factors.
Other Examples:
Let me address a question someone asked in an Internet video forum: "[I] have a video projector with 1280 by 1024 resolution non-interlaced. Will it display a 1080i HDTV picture?"
I would have to say "no". The video device would need built in circuits including video memory buffers whose specific purpose is to take the interlaced input and feed it to the non-interlaced image drawing circuits. Without such conversion circuits, the video device would probably switch to 540p mode, keying in on the scanning frequency of the input signal but treating each incoming field as an entire frame. The resulting 540 line image would have half the vertical resolution of the original 1080 line image.
Even if it did the scan conversion correctly, the video device is rated for 1024 scan lines, not 1080. It would either discard the first 14 and the last 14 lines of each field, or reformat (downconvert; scale) the incoming 1080 lines to be 1024 lines. The former method would probably give a better quality picture.
An inexpensive method of displaying NTSC video on a progressive scan monitor such as a VGA computer monitor uses interlace reconstruction. This writer does not know of any products that utilized this technique but one advantage is not needing a field's or frame's worth of video memory.
The standard 640 pixel wide by 480 pixel/scan line high VGA scan rate is double the NTSC scan rate. Using re-interlacing the monitor paints on the screen: odd line, black line, odd line, black line, etc. for the odd field and then: black line, even line, black line, even line, and so on for the even field. Thus the NTSC signal is fed into the monitor with a minimum of processing while the monitor is scanning at its normal rate. Technically the monitor is painting a 480p image but since every other line is black the visual result, if done correctly, is a 480i image identical to the original source. Only two lines' worth of video memory are required.
What is a Sharpness Control?
The sharpness control is not a focus control; it does not change the electron beam spot size. The sharpness control gives the illusion of higher horizontal resolution by electronically altering the video signal so big light to dark transitions appear more pronounced. Technically the frequency response of the video circuits are altered so that higher video frequencies (fine horizontal details) are accentuated.
In a sense the sharpness control is a gimmick. It is found even on TV sets without comb filters and therefore having no more than 260 lines of horizontal resolution to begin with. Like bass and treble audio controls, the sharpness control should be first set to its "off" or "neutral" position and then optionally carefully adjusted to suit the viewer's personal desires.
Unfortunately the "off" position of the sharpness control is not easy to find. It varies from one TV set manufacturer to another; on some sets it is at the "left" while on other sets it is near the "middle" with artificial further softening of the picture as the control is moved "to the left". Finding the "off" position of the sharpness control may require the help of a test video such as Video Essentials.
Often, increasing "sharpness" introduces or accentuates ringing, ghost edges close to and to the right of dark to light or light to dark transitions. The electron beam as it changes from strong to weak or vice versa bounces a few times. The cause may be anywhere in the electronics.
Line Doubling and Related Terminology
480i stands for a 480 scan line high picture presented as two 240 line interlaced fields (not counting the black lines 481 through 525 half of which go with each field). 480i can also stand for digitally formatted video pictures that don't possess scan lines 481-525.
480p stands for a 480 scan line high progressively scanned picture. The VGA computer monitor standard 640 by 480 pixel resolution is an example of 480p.
240p can refer to a picture originally intended to be 480i but where the even lines landed on top of the odd lines, thereby cutting the vertical resolution in half.
The highest resolution official U.S. high definition TV format is 1080i; interlaced 540 line fields with 1920 pixels on a line. Next comes 720p, 720 progressively scanned lines making up the picture and with 1280 pixels per scan line.
A "field" consists of just the even lines or just the odd lines (half of the total picture) during interlaced scanning.
All parts (c) copyright 1998-2006, Allan W. Jayne, Jr. unless otherwise noted or other origin stated.
We regret that we have not used or evaluated the many video processors
on the market and therefore cannot make any recommendations.