IMDb DVD Quality Reviews: Technical Background

by Michel Hafner (IMDb) and Peter W. Simeon (IMDb), technical consultant: John G. DeGroof

General remarks

All IMDb reviews of technical DVD quality are made with one of our two reference systems. The viewing distance is approximately two times screen height (1.33 : 1 screen aspect ratio). All systems have been properly calibrated for NTSC with the Video Essentials DVD. For PAL, unfortunately, there are currently no comparable DVDs for calibration available, so we calibrated the systems using DVDs of the same film made from the same master for PAL and NTSC by making the PAL picture look like the NTSC picture. When we refer to normal viewing conditions we mean a setup that has been calibrated that way (which includes no ambient light for a screen/projector system) and a viewing distance of close to two times screen height.

Our current two reference systems are as follows:

Video:

• DVD Player: Sony DVP-S7700
• Video Projector: Barco Cinemax high end 9 inch CRT
• Video Processor: Faroudja VP 401U, NTSC/PAL line quadrupler
• 4:3 Stewart Screen StudioTek 130, microperforated, THX certified, 2.4 m wide
• Cable: Monster M1000V, YUV connection for DVD player, RGB for video projector

• Surround Processor: Meridian 861
• Speakers: 5 x Campana 3 (high end direct digital active speakers, 24 bit, 48 Khz)

Video:

• DVD Player: Sony DVP-S7000
• Video Projector: Barco 1209S high end 9 inch CRT
• Video Processor: Faroudja VP 3000, NTSC/PAL line quadrupler
• 4:3 Screen
• Cable: Belden, YUV connection for DVD player, RGB for video projector

• Surround Processor: Meridian 861
• Speakers: 5 x Campana 3 (high end direct digital active speakers, 24 bit, 48 Khz)

Before we write a DVD review we watch the whole DVD at least once from start to finish. If necessary we watch it partially or completely a second or more times to make sure that we got it right. The time we can invest per DVD is limited though, since providing DVD reviews is just one of many tasks we have to perform. That is also the reason why we do not systematically review every new DVD that is coming out, be it in region 1, 2 or any other region. There is simply no time to do that. We have to select, and select we do. We try to review what is popular and current, what promises to be at least of acceptable quality, and also what is of high historical importance and/or has great artistic merits. Finally, from time to time, we review what we personally like to see, even if it may appeal to a minority only. We do not review DVDs from 'junk labels' that test the lower bounds of quality achievable on DVD (if such a bound exists), and we do not review X rated material . We also do not review DVDs that present the film with a pan-and-scan or full frame version only, if the theatrical presentation was different. We do not support this kind of tampering. Finally, when a DVD offers a widescreen version and a pan-and-scan version, we review the widescreen version only, since we don't have time to waste on mutilated versions of the films in question. Experience shows that the pan-and-scan version has usually lower image quality than the widescreen version. Another reason to avoid it. But also a warning, that a verdict of excellence for one version does not necessarily imply that the other is equally good!

If your impression of the technical quality of a DVD is very different from ours, please keep in mind that there are many factors that influence the apparent image and sound quality of a DVD, played on a specific system.

The most important are:

• Quality and characteristics of the display system:
  If the setup for playing a DVD is not as good or better than the DVD itself, you can never be sure whether you are seeing a problem of your setup, a problem with the DVD, or a mixture of both. The quality of the setup is defined by the quality of your DVD player, connecting cables, video processor and display device (TV set or video projector plus screen). It's important what kind of display technology you are using. The sought after film look of DVD requires that you use a video processor that produces progressive images out of the interlaced signal coming from the DVD player, or a progressive DVD player. The latter type of players is just entering the market now (end of 1999). It also requires that you use display and sound equipment that is capable of fully resolving all image detail stored on a DVD, including correct color and grayscale rendition, taking advantage of 16:9 enhanced DVDs and 5 channel digital sound. The current state of the art display devices for large screen home cinema are still high end CRT projectors whose image quality is unsurpassed when driven by a suitable line multiplier. If you want correct color and grayscale, deep blacks, excellent shadow detail, and images free of pixelisation, projector related motion artifacts (such as smearing and ghosting) and/or a more or less fine grid or a gray veil over everything, you need a good quality CRT projector. LCD and micromirror/DLP technology are currently inferior and not suitable for home cinema, if you strive for the utmost quality. LCD projectors, especially, have severe limitations and do not offer the full quality available from the DVD medium. LCD and DLP projectors are making progress though and especially the latter are likely to catch up with CRTs sometime in the not so distant future. Also very important is the quality of your video processor. Without a decent line multiplier (doubler, tripler, quadrupler, variable scanrate device) that gives you correct progressive frames as they are on the original film itself, you can not get really good wall size video projection. Line multipliers that come built into a projector are almost always of moderate or bad quality. If you use a regular TV set instead, you will not get a true home cinema experience, since the image is simply too small (even for the larger models) and its interlaced nature creates motion artifacts that you cannot blame on the DVD (again, if a doubler is built in it's almost certain to be of mediocre or bad quality). Our two reference systems work with HDTV resolution and can fully resolve the quality of DVD video in all its aspects. If a DVD looks good on our setup, it looks good on any properly calibrated setup with correct grayscale rendition. The other way around is not necessarily true though, as many (rave) reviews on the web, made with lesser quality systems, quite clearly show.
  
  
• DVD mastering process:
  DVDs are mastered to look good (well, leaving aside the junk labels and titles) on interlaced reference studio monitors of moderate size, often with an average consumer monitor as back up to represent what Joe Average Consumer will see in his home. They are not specifically mastered to look good on ultra bright LCD or DLP projectors that can't provide real blacks, and they are not specifically optimized for progressive displays or real large wall projection (2m wide and more), although excellent quality DVDs will look fine on any kind of display. This means that, although the expert eye can see everything that is to be seen on an interlaced studio monitor, some artifacts are less pronounced and/or harder to see and/or less objectionable on such a monitor than on a very bright display with bad black level or a progressive display or when projected on a large screen. Therefore some artifacts can look (much) worse on these systems (even when they are properly calibrated within their technical limits) than on the studio monitors used for DVD mastering. A good CRT projector or CRT monitor on the other hand creates pictures that are rather close to what a studio monitor provides, so it usually will better show you what the people at the DVD lab saw and optimized for and therefore it will often give a you a better looking picture for this reason alone. Some artifacts that are relevant in this context are:
  • Compression artifacts: Can be better visible on very bright displays with bad black level since image content that is supposed to be hidden in the dark is now too bright.
  • Overenhanced edges: Are (much) more objectionable on large displays than on small displays. On the former they create well visible halos and ringing artifacts around contours, on the latter they are harder to see (because people sit farther away relative to the screen height) while the picture looks sharp and pleasing. To learn more about edge enhancement on DVD click here.
  • Digital noise reduction artifacts: Are better visible on progressive displays (because the veil of interlace artifacts has been removed (film source and pulldown reversal) or reduced (deinterlaced video source)) and large displays (because people sit closer in comparison to the screen height). Also, if a scaler is used the less scaling artifacts it creates the better you can see the noise reduction artifacts as on DVD. You can see noise reduction artifacts best on progressively projected wall size images, using a projector with correct reverse pull down capability and no motion and scaling artifacts of its own. To learn more about digital noise reduction (artifacts) click here.
  • Grain and noise: Better visible on progressive displays and displays with little noise of their own and no projector related motion artifacts.
  
  
• Calibration of the system:
  If your setup is not properly calibrated, you may not be able to see a specific quality problem, or you may see it to a much greater extent than would be the case with a properly calibrated setup. This includes the environmental conditions (ambient light and sound, reflections of light from walls on the display device, acoustics of your room etc.). We recommend that you calibrate your setup with the Video Essentials DVD or some similar DVD to make sure that contrast, brightness, sharpness and color are properly adjusted.
  
  
• Viewing distance and projection speed:
  If your viewing distance is much more or less than two times the screen height you may not see some problems or they might look to you much worse than with our standard viewing distance. If you distance yourself enough from your screen, all DVDs will start to look good, since you can no longer resolve crucial image detail. If you move unreasonably close ( <= about one screen height) you will always see a fuzzy picture and various artifacts that would not be present on, for example, a projected 35mm slide. Be aware that DVD does not offer the kind of resolution that 35mm offers. Sitting closer than about 1.5 screen heights is not recommended independent of the quality of your setup or the DVD watched. Be aware also that if you watch your movies on a more or less small TV screen, you are likely to sit much further away than two screen heights. And you will not see a lot of the problems we report. You will also not have a home cinema experience, but a home TV experience. There is nothing wrong with that, but keep in mind that our reviews are made on and for home cinema equipment in the first place. Finally, we evaluate image quality at the regular projection speed of 30 full frames (25 for PAL) per second, and not when stepping through individual frames by hand. Some artifacts are much more apparent when stepping through frames, others less.
  
  
• Personal preferences, measurement/evaluation problem:
  Individuals have different opinions about what the criteria for image quality should be and how important they are. Some people might refuse to watch a DVD simply because it's not in color. Others might object to compression artifacts but not be bothered by (slightly) fuzzy pictures. Still others might detest scratches and speckles but not care about color bleeding. Most criteria are ambiguous in nature or undecidable, maybe technically/theoretically well defined, but difficult or impossible to measure/evaluate in practice,, unless you have special knowledge and/or equipment:
  
  • What is the correct hue of red in that scene as the director of photography intended it to be?
  • How much of this noise is coming from the film master, the video transfer, the MPEG compression? How much is due to poor mastering and how much is there because the director wanted a 'grainy look'?
  • Is this compression artifact slight or severe, distracting or not distracting, easily detectable or hard to see?
  • ...
  There are often no unique and definitive answers. It depends on your preferences, experience and special knowledge, and finally subjective decisions. So, if a DVD looks wonderful to you, let nobody ruin the day with claims, that it is of low quality, one of the worst and so on. You are right for your current circumstances and current preferences. These might change, so might your judgement. When we are able to be of any help here with our technical DVD reviews, we have met our goal.

Image Quality Categories and Evaluations

• master source quality
• sharpness
• color rendition
• contrast
• video artifacts
• compression quality
• noise/grain

Video Average

Let's now have a closer look at the different image quality criteria.

Master Source Quality

• film source: speckles, scratches, nicks, scars, tears, missing frames, other artifacts typical of poor handling or storage of film stock, plus image steadiness (can be a problem of the source itself or the teleciné, we evaluate it with the source).
• video source: drop outs, time base errors, chroma noise

Sharpness

• Master elements used are not as sharp as DVD itself.
• Loss of sharpness during the teleciné transfer.
• Loss of sharpness due to digital filtering (noise and grain suppression, vertical detail attenuation to minimise flicker on interlaced displays, low pass filtering to remove high frequencies MPEG compression can't handle with available bit rate etc.)
• Loss of sharpness due to lossy MPEG compression, usually made worse by rounding errors in the MPEG hardware.
• Unavoidable loss of sharpness due to the 4:2:0 format used for DVDs. Color difference information is stored only for every second pixel on every second scanline.

Color Rendition

• natural skin tones where they can reasonably be expected, natural looking colors for shots of landscapes, animals etc.
• fully saturated colors, where they can be expected
• absence of general tints, where there very likely shouldn't be any.

Contrast

• judging the amount of shadow detail where shadow detail can be expected (sometimes a scene is shot in such a way that there is no shadow detail, so strong silhouette effects are achieved. Asking here for shadow detail would be silly)
• good use of the full contrast range
• linear gray scale
• absence of hard clipping at the upper end (white) or lower end (black)

Compression quality

• blocking artifacts: The images look as if they consist of tiny blocks.
• static and non static noise patterns: The removal of high frequency content by the compression adds general noise to the images that can be easily seen in severe cases. A special kind of compression noise that is called mosquito noise looks like mosquitos around an object, usually a sharp edge as in letters of film credits. In addition the compression can cause stationary noise patterns that hover over moving image parts and look as if you were projecting on a dirty screen (the same effect can be caused by a bad teleciné whose phosphor is visible, so called tube face, which is evaluated under video artifacts).
• image break up: Some MPEG encoders have a tendency to break rigid objects (such as a background or complete scene) into different parts, which they then move around independently. The general scene can be non-moving, in which case some objects are suddenly twitching around, or it can be more or less moving, in which case the break up produces image jitter or unnatural looking motion within rigid objects.
• loss of detail in moving or stationary image parts (color and structure)
• oscillating image quality concerning sharpness, image stability etc., going from good to bad and back to good in rapid succession (so-called I-frame pulsing)

Video Artifacts

• aliasing: The appearance of jagged lines and edges and/or a rough and jittery look of textures where there should be a smooth but sharp impression instead. Aliasing artifacts are generated when fine image detail that cannot be correctly resolved with the resolution of the video master is not removed before the master is created. This happens if the film element is too sharp for the resolution used in the teleciné, or if an aliasing free digital high resolution master is not properly downsampled to NTSC/PAL DVD resolution. Aliasing is very widespread, and there are hardly any DVDs that have none, but again the degree of aliasing varies considerably between different titles. Also, aliasing is much more evident on wall-size progressive scan images than on a (small) TV with interlaced images that suffer from jagged edges anyway due to their interlaced nature.
• overenhanced edges/ringing artifacts: To make an image look sharper one can boost the high frequencies in the image. Unfortunately this also leads to ringing artifacts around edges of objects and white outlines around contours, if it's not done with the necessary caution. DVDs without ringing artifacts are very rare, although the visibility of these can vary enormously from title to title. In addition overenhanced edges can look good on small TVs but awful when projected on a wall size screen. We evaluate DVDs when projected on a big screen. Also it's important to know that ringing artifacts are generated to some extent by the digital-analogue conversion in the DVD player, in low quality cables, and within the projector and video processor.
• color banding: Color banding happens if smooth color gradations cannot be correctly rendered and are instead displayed as more or less constant color bands that are clearly distinct from each other. Although a DVD does not have the color resolution of film, it should not exhibit any color banding if it has been properly mastered.
• tube face: Visibility of the phosphor structure of the teleciné in form of a static noise pattern that hovers over images and looks like your screen is dirty.
• comb filter artifacts: The video signal on a DVD is stored in digital component format (YUV). Comb filters are designed to separate the chrominance (UV) and luminance parts (Y) of a composite signal (as used on LD and VHS tape). Since DVD does not work with composite signals it does not need a comb filter, one would think. That's true. Comb filters are needed during DVD mastering when the video master is in composite format instead of the much better quality component format, and needs to be split up for further processing. Modern telecinés use the component format right from the start and the image signal remains in component form throughout the whole DVD mastering process. When a DVD is mastered from an older composite transfer though that was made several years ago for LD/VHS, or if the original source is composite video, then the use of comb filters is unavoidable. Since these filters never perfectly separate the chrominance and luminance signals you get comb filter artifacts such as rainbow patterns (high frequency luminance detail is interpreted as color).
• chroma noise: This is noise in the U and V channel on the DVD not due to the original source or MPEG compression.
• chroma delay: When the Y and UV channels are not properly aligned edges of objects are blurred and get unwanted colored halos.
• color bleeding: Especially highly saturated colors can bleed over the edges of objects into to the surrounding neighborhood. To some degree this is unavoidable on DVD since the U and V channels are only stored with half resolution compared to the Y channel. This alone is rarely distracting. But in addition to that there is often more bleeding due to suboptimal masters and processing.
• digital noise removal artifacts: Modern video processing equipment allows the removal or attenuation of film grain and other forms of random noise from the digital video images. This is achieved by computing averaged pixel values using succeeding frames in the video stream. As long as there is no movement in the frames, and the camera is static, this averaging works very well. But if there is movement the averaging produces artifacts, unless the motion is precisely determined and undone before averaging. This is not possible with real-time equipment. Therefore digital noise removal tends to produce artifacts as soon as it's applied on moving image parts. The artifacts are defined as sudden loss of fine image detail when the movement starts and reappearance of the detail when it stops. Alternatively fine image detail can be slightly flickering when textures are slowly moving or showing an unnatural appearance that is not caused by general noise. Additionally there can be ghost contours around fast moving objects. Finally noise reduction algorithms can break up rigid image parts and move the pieces around individually, giving the whole images a jittery look. Whether the artifacts are visible/distracting or not depends on the sophistication of the noise removal algorithm, the experience of the people using it, and the speed and detail of moving image parts. There exist also algorithms designed to automatically remove random scratches and speckles from the frames. These algorithms must distinguish between the unwanted scratches/speckles and the authentic image information. Again something that cannot be done in real time with 100% accuracy in all circumstances. If the algorithm errs and tries to remove image parts that should not be removed, you get scratch removal artifacts.

Noise/Grain