Super Resolution vs. Reality Creation vs. Darbee Darblet
Cine4Home sharpness showdown
The purpose behind the introduction of “Full HD” a few years ago is explained by the name itself: “high definition” means nothing other than a highly detailed, sharp picture. And indeed, with around four times the resolution of conventional PAL (standard definition), HDTV brought a huge leap in detail reproduction that anyone could see.
And yet the optimum has not been reached, and the limits of healthy visual acuity have not been exhausted, especially with large-screen projection: manufacturers are hard at work on “Ultra HD” resolution, which is already almost standard on TVs – only the content industry is lagging behind.
Speaking of the content industry: many big-screen fans rightly complain that Full HD resolution alone is not the limiting factor for sharpness; rather, many films on Blu-ray and TV broadcasts do not use the full resolution potential at all, accepting losses in favour of lower bandwidth. In practice, the maximum possible picture quality of Full HD is rarely achieved. So Full HD “can do even better”.
The consumer electronics industry has recognised this potential too and has for years offered intelligent signal processors that enhance Full HD sources after the fact – first as external scalers, and for some time now as a “free extra” in high-quality TVs and projectors. Every manufacturer gives its technology a catchy marketing name: Super Resolution, Reality Creation, Detail Clarity and so on.
In principle, the algorithms are designed to define edges more sharply, add resolution and boost contrast, as these are all picture aspects our eyes subjectively perceive as greater sharpness. However, they also carry the risk of digital distortion and over-sharpening.
In this special, we compare three of the best-known systems currently on the market under real-world conditions: the “Super Resolution” of the Epson EH-TW9200, the “Reality Creation” of the Sony VPL-HW55 and, as an external solution for all other projectors, the “Darblet” processor from Darbee. What picture improvements does each system deliver, what side effects do they have, which system is most convincing, and are such systems worth recommending at all? We answer all these questions on the following pages…
Our comparison took place at the renowned projector specialist “Heimkino-Boutique”: owner Gunnar Witting keeps all models ready for demonstration and can personally show you the comparison presented here at any time under ideal home cinema conditions – so a visit is well worthwhile!
1. How it works and possible side effects
Before we compare, we would like to explain the technical basics: how exactly can the sharpness of a video image be increased without raising the native resolution?
1.1 More sharpness by manipulating brightness?
To explain the principle of sharpness enhancement, we use a well-known example. As our starting point, we look at an image with elements of varying detail – the familiar bridge from the HQV test disc:
Original image, as stored on the signal source
The panorama looks detailed and convincing, but it has untapped sharpness potential, as the following processed version shows:
Digitally sharpened
Although this is exactly the same image content at the same resolution, the processed version looks much sharper. The reason quickly becomes clear when we magnify small details. Let us take a closer look at the steps on the left, first in the softer original version:
Magnified, the steps look like slanted light grey -> dark grey -> light grey bars. Each step is clearly defined, but a kind of veil hangs over the image. Now let us look at the “sharpened” steps in close-up:
On closer inspection, you notice that the dynamic range of the steps has been stretched. This means the dark gaps have been darkened (darker grey), while the light steps have been brightened (lighter grey). This increases the contrast between the details and strengthens their separation, so the image looks subjectively sharper.
But the first side effects are also apparent: the harder contrast transitions also emphasise the stair-stepping, and the magnified image becomes more pixelated and prone to artefacts. Fortunately, the native resolution of Full HD projectors is so high that such pixel-sized “aliasing” effects are not very significant. Even so, this hides an annoying side effect that should definitely not be overdone. In the next example, we increase the added sharpness even further to reveal another danger:
Now the light and dark elements are stretched even further apart, and the steps consequently look even sharper. But you notice that some of the light steps have already reached the maximum level (white); the yellow/beige character of the original is almost completely lost, and the steps look paler, overexposed and almost artificial.
1.2 Possible artefacts / requirements for a sharpening algorithm
We now know that subsequent sharpness enhancement is a kind of “pixel-based gamma correction”: if a lighter element sits next to a darker one, the contrast is increased by stretching (brightening + darkening) so that the separation is brought out more strongly, which our eyes interpret as “sharper”.
Contrast at pixel transitions is increased, strengthening the impression of sharpness
You can see that no real detail is added (how could it be, if it is not in the image signal?); the existing details are merely emphasised more strongly. This is a simple and effective method, at least in theory, because it carries various risks:
1.2.1 Overexposure
We have already identified the first side effect above: if you stretch the pixel-based contrast range too far, you quickly reach the white range, and the image looks artificial, overexposed and pale.
In this sample image, the overexposure is particularly visible on the window frames and columns. They are already very bright in the original (top) and are pushed into white clipping by the pixel-based contrast boost:
The result (top) makes the photo look almost embossed. If you overdo the sharpening, colourful elements suffer too, as the following example shows:
Top: original, bottom: over-sharpened
Here too we are dealing with clipping, which seems to settle on every detail like white dust. When assessing sharpening algorithms, it is essential to watch out for artefacts like these.
1.2.2. Aliasing
Aliasing is the technical term for resolution-related stair-stepping. Gamers in particular will be familiar with “anti-aliasing” filters, which ensure that fine diagonal lines and elements in the image look natural rather than jagged. Sharpness enhancement can also cause aliasing as an unpleasant side effect. Consider the following image, which could appear in any feature film:
In the foreground is a city street with the usual abundance of markings. In the original above, these markings look convincing, even though the resolution is already reaching its limits with the slanted white diagonals. However, a certain softness prevents strong stair-stepping. If we sharpen this softness away, it looks like this:
The image certainly looks much sharper now, but the contrast boost makes the road markings look unnaturally jagged and digitally rasterised. In motion (e.g. during a camera pan), such artefacts are amplified further and can result in line flicker. Intelligent sharpness enhancement should avoid aliasing artefacts.
1.2.3 Increased noise
Everyone knows the effect from photography or has often seen it in films at the (home) cinema: uniform coloured or grey areas do not look truly even on the screen but are speckled with thousands of tiny dots. These move from frame to frame, creating visual noise. So this is not a digital artefact of the projector, but an artefact of the analogue celluloid medium.
Opinions are divided on noise, as the introduction of Blu-ray and various compression algorithms showed: the “conservative” camp among filmmakers and studios believes that visible “film grain” is an artefact worth preserving that should not be smoothed out during mastering. Others prefer the smoother, “cleaned-up” look. Be that as it may, most DVDs and Blu-rays still show visible film grain today. Let us look at it in a magnified single frame:
Single-colour areas are noisy up close
On closer inspection, the supposedly uniform area appears speckled, with many dark and light spots and dots scattered around the “target colour”. These move from frame to frame, creating an even noise.
As already explained, sharpness is achieved by increasing the contrast between pixels. For our “noisy” area, this means the contrast within the speckle is increased: the darker spots/pixels are darkened and the lighter ones brightened further. As a result, the noise becomes much more pronounced:
Sharpness enhancement amplifies noise
A sharpening algorithm should therefore either detect such noisy areas or work so moderately that the noise is not significantly increased – certainly no easy task.
1.2.4 Double contours
Conventional sharpness settings on TVs and projectors in particular are known for these artefacts: if you turn the sharpness control up too far, unsightly double contours soon appear in the image, also known as “ringing”.
Such contours mainly appear next to strong contrast transitions from light to dark or vice versa. They are especially obvious against the sky in backlit shots (see example above).
It goes without saying that powerful, modern sharpness enhancement should not create such double contours. And yet there is a problem: the source material or signal source often already shows minimal ringing – barely visible double contours around sharply defined objects. With conventional sharpness these go unnoticed, but subsequent sharpening raises the contrast at pixel level and, as with noise, this artefact then becomes clearly visible.
Sharpness enhancement emphasises double contours
Here we enter an area that is very difficult to get around through “circuit intelligence”. Full object recognition is still not possible.
1.2.5 Intentional blur
Raising the sharpness evenly across every image is not necessarily the right approach, because sharper is not necessarily better. One particular point that many home cinema fans and “sharpness fanatics” overlook: a cinema image is not meant to be sharp in every area (depending on composition)! That may sound strange at first, yet blur is the most natural thing in the world:
Our eyes are designed to focus on only one distance range at a time. If you concentrate on one object, only that object (and others at the same distance) appears sharp; everything else is rather blurred and therefore receives less attention and importance in our perception. The same applies to a camera.
And good directors make use of exactly these properties of human perception: by clearly defining the focal plane, they influence what the viewer concentrates on in the image and which elements receive more attention than others. This is essential to guide the viewer’s gaze, because the image is two-dimensional (the screen is flat and always sits on a single focal plane of the eye), so the brain has no visual cues for “important action” other than sharpness:
Original…
The sample image above makes this clear: the protagonist (Johnny Depp) is at the centre of the action and is therefore shown in sharp focus, while the background and the wall on the right appear blurred and unimportant. The different planes of focus give the image significantly more spatial depth.
…conventionally sharpened…
If you sharpen the entire image using conventional methods (image above), it loses this spatial depth: wall and protagonist sit on the same plane, the eye is distracted and the image looks flat. The following example makes this even clearer:
…distorted by incorrect sharpness enhancement.
In this example, the wall is sharper than the actor and therefore moves into the foreground spatially, even though it is of no importance to the action.
Because of these important aspects, intelligent sharpness enhancement must distinguish between the focal plane/important elements and “side scenes”. In the image above, for example, it only makes sense to sharpen Johnny Depp’s face, not the wall or the background. This is the only way to preserve the original composition – and the targeted sharpening even adds depth:
Top: original, bottom: sharpened only in important
areas
Image depth increases without distorting the composition
This example shows how an intelligent, complex sharpening algorithm can visibly improve picture quality without disturbing the composition. The actor stands out even more clearly, while the wall and background remain untouched and “unimportant”. The image looks even more three-dimensional and appealing.
1.3 Requirements for a sharpening algorithm
Our overview of the possible problems has shown that increasing the sharpness of a moving film image is not that simple. Overdo it and you quickly provoke artefacts (aliasing, clipping), amplify existing ones (noise, double contours) or distort the intended composition (focal planes). Are the two sharpening technologies from Epson and Sony able to take all these aspects into account?
2. The test: real-world comparison of each system
Having covered the basic principles and possible side effects in the previous chapter, we now want to apply this knowledge to the three competing systems. To do so, we choose two well-known film scenes and examine the characteristics of the sharpening in detail:
2.1 Epson Super Resolution
Among projector makers, Epson is the pioneer of subsequent sharpness enhancement. The classic sharpening algorithm is therefore the well-known “Super Resolution”, which has been very popular in the top models of market leader Epson for years. It comes from NEC and is generated by a dedicated processor, the “NEC 9280”.
The NEC sharpening chip on the main board
Further research on the websites of electronics group NEC confirms that this is indeed a signal processor whose task is to visibly improve picture sharpness. It is part of the “Neo Clear Resolution” family.
And lo and behold: “Single Frame Super Resolution” shows that the name does not come from Epson either, but from NEC. This also explains why Super Resolution can now be found in more and more projectors and TVs.
“Single Frame” means that the image analysis is always based on a single frame. Including subsequent frames in the calculation would undoubtedly achieve even greater accuracy, but this currently still exceeds the available processing power.
We illustrate how the algorithm works with a flow chart showing the individual steps and their sequence:
How “Super Resolution” works
In the first step, the incoming image is analysed for blur. Wherever the processor detects sharpness potential, small details are to be defined more sharply using “Mixed Color Separation” technology. Another English feature name – so what lies behind it? When you scale an image to a higher resolution (from PAL to HD, for example), edges are often softened by inserting averaged colour values as missing pixels:
This scaling method makes the result look very natural and fine, but maximum sharpness is not achieved. This is where colour separation comes in: the algorithm is designed to detect and remove such colour and brightness transitions without distorting shapes and structures:
In the final step, the newly calculated “Super Resolution” image is compared with the original, processed once more and finally leaves the signal path.
This brings us to the main focus of NEC Super Resolution: it was originally designed to recover more sharpness from conventional SD material (PAL/NTSC) after scaling, but the circuit is also meant to add sharpness to HD material using the pixel-based gamma method we explained at the beginning of this special. It is also clear, however, that this circuit offers a great deal of potential when upscaling to 4K. The Super Resolution principle forms the basis of all more advanced techniques, including the following…
So much for the theory – now for the practical test. We look at a scene from “Batman” that every film fan knows: the last quiet seconds before the Joker strikes…
The screenshot above shows the scene in its native Blu-ray form without any post-processing. Our choice looks surprisingly unspectacular, as this scene does not exactly “cry out” for extra sharpening. Even so, it offers considerable potential as well as pitfalls that reveal the quality differences between the algorithms.
Let us first look at the Joker’s dark trousers: against the light street, there is no significant lack of sharpness in the original. In the macro shot, however, you can see that the protagonist by no means stands out “black on white” from the background. The edges are slightly softened with shades of grey – and not without reason: these “micro grey steps” stop the limited Full HD resolution from becoming too obvious and reduce stair-stepping and the “aliasing” artefacts explained above, which would make the image look artificially digital. So a compromise between detail and naturalness has been struck here.
But this mastering is not perfect: if you look closely, a slight double contour next to the legs becomes visible – the unsightly and unwanted “ringing”.
So how does Epson’s “Super Resolution” handle this detail? The intensity of its processing can be set in 5 levels. We photographed each level, and the result speaks for itself:
Level 1…
…2…
…3…
…4…
…5…
Indeed, the subjective sharpness is increased dramatically and the algorithm works exactly as the manufacturer intends, but whether picture quality is actually improved is debatable.
On the one hand, Super Resolution consistently removes all the “softening grey steps”, increasing the contrast between object and background and thus the subjective sharpness. Unfortunately, though, slanted edges become so “jagged” that the unwanted aliasing artefacts become all too obvious. In the moving picture they are even more noticeable as “edge flicker”; the image looks poorly digitised and the limits of Full HD resolution become apparent.
Original (top) & Epson Super Resolution (bottom)
At the same time, the algorithm does not recognise the subtle “ringing” for what it is and tries to sharpen the double contours too. This amplifies the artefact further and makes the image look even more artificial. If you prefer a natural picture, only the first two, weaker Super Resolution levels are usable: they effectively emphasise sharpness without the annoying side effects being too noticeable from a normal viewing distance. We can only recommend levels 3 to 5 to sharpness fetishists who “don’t mind” artefacts.
Overall, the scene is sharpened everywhere, including where it should not be: the houses in the background are deliberately very soft to create greater depth and push the protagonist even further into the foreground (see the example in chapter 1). The composition intended by the filmmakers is compromised.
For the second scene, we choose our classic: St Mark’s Square from Casino Royale. Here every algorithm can show how it enhances an image that is already highly detailed. The original shows fine structures and numerous details only one or two pixels in size.
In the first two levels, Epson Super Resolution succeeds in bringing out the fine structures more clearly by increasing contrast, thereby boosting subjective sharpness. In levels 3 to 5, however, the disadvantages once again outweigh the benefits:
The building wall on the left is overexposed and looks punched out, like a Photoshop emboss filter. The crowd in front of St Mark’s Basilica also loses its colour and turns into white specks. This can hardly be called an improvement in quality.
Much sharper, but with many side effects
At the same time, the noise, which is already above average in this film, is amplified further and naturalness is reduced even more. Overall result: much “sharper”, but very unnatural.
Note:
All the above results refer to the latest firmware version of the Epson EH-TW9200, in which Super Resolution is emphasised particularly strongly to give the untrained eye more of a “wow effect”. However, we also had an Epson with the previous firmware version available, on which Super Resolution can be dosed much more finely. Here is the direct comparison:
Left: new Epson firmware, right: old firmware
As you can see, the “old” Super Resolution works more subtly and less conspicuously, but strikes a much better balance between sharpness gain and naturalness. We therefore recommend not(!) upgrading to the latest firmware.
2.2 Sony Reality Creation
Behind Sony’s own marketing term lies a genuinely Sony-developed technology. Several processors are responsible for intelligent detail adjustment: the X-Reality processor handles artefact-suppressing algorithms such as intelligent noise reduction, while others improve detail reproduction according to the “Super Resolution” principle (see above). Sony likes to present its algorithm as a large “texture database” that recognises details independently and then inserts matching patterns by itself.
However, this description belongs in the “marketing hype” category, because in practice there are no thousands of graphics stored, as Sony would have us believe. Nevertheless, the algorithm is very complex:
Experienced home cinema experts know that the mathematical fast Fourier transform is a key part of digital compression for picture and sound and is what makes it possible in the first place. In video, it can be used to convert picture information into its frequency ranges, enabling further analysis:
This frequency conversion makes it possible, for example, to distinguish image areas with fine detail (high frequencies) from those with little detail (low frequency). In the example above, these are marked in different colours. Note that simple object edges are not classed as high detail – only fine structures are.
Sony’s engineers carried out this FFT frequency analysis on countless images while developing their picture processors and collected typical, i.e. recurring, characteristics in a database. The database should therefore not be understood as a simple “graphics store” containing various patterns, but rather as a collection of mathematical frequency conversions (conversion database). Sony’s many years of experience in 4K capture and mastering naturally help here.
So how effective is this system? That depends mainly on the source material: the “sharper” the Full HD (or SD) source was mastered, the more “traces” the Reality Creation algorithm detects and the more effectively it can apply its resolution-enhancing algorithms.
Reality Creation copes best with Full HD material that was derived (downscaled) from a 4K master and compressed. Here the algorithm detects the typical 4K -> 2K scaling artefacts, and image areas with high frequencies come to light.
This sample image covers all frequency ranges: the gate’s shadow and the back of the board are large areas (low frequency, blue), the masonry and grass show finer details (medium frequencies, yellow/green) and the inscriptions on the gravestones show very fine details (high frequency, red). In other words, in this sample image Reality Creation works particularly effectively on the walls and the inscriptions.
But there are also Full HD masters where Reality Creation cannot work such wonders: this is usually the case when the film was originally scanned in 2K only and fine details had already been flattened in frequency during compression.
This second example shows such a master: even in fine structures such as the hair, facial skin or the blazer, compression has already “filtered out” all high frequencies – think of it as a band filter to make compression more efficient. With material like this, Reality Creation does manage to increase subjective sharpness through the typical “Super Resolution” pixel analysis, but no more real detail is brought out.
So far, so good: the algorithm is able to recognise finely resolved image content and potentially make further use of it. In actual display, however, there is a problem: a conventional Full HD projector does not offer much resolution headroom over the source in which to add picture information to the detected details. The algorithm must therefore limit itself to minimising artefacts, either via pixel gamma or conversion.
So is it all just marketing hype with no real benefit? Let us simply take a look again:
Top: original, bottom: Sony Reality Creation
At first glance, the Sony HW55 also achieves a noticeable increase in sharpness. Details on the bag and hair are emphasised more strongly but do not look too digital. However, the deliberately soft background is clearly sharpened here and picture noise on the street is emphasised.
Reality Creation is also unable to recognise the unwanted ringing in the original and amplifies the annoying double contours so that they become visible from a normal viewing distance – though not to such a crude degree as the Epson. Aliasing and jagged edges, on the other hand, are not provoked; here Reality Creation is smarter.
Top: original, bottom: Sony Reality Creation
The St Mark’s Square scene also shows the pros and cons clearly: sharpness is increased significantly without the punched-out, distorted look of the Epson. The colour of small details is preserved, too.
Top: original, bottom: Sony Reality Creation
The film’s noise, however, is amplified considerably and clearly spoils the otherwise good picture – and Reality Creation’s separately adjustable noise reducer does not help much. Based on all our tests, you should not set Reality Creation higher than 30 (the scale goes up to 100) if you want to preserve a natural look.
2.3 Darbee Darblet
The Darbee Darblet is a small box that is inserted into the HDMI signal path. At around €250 it costs more than many a Blu-ray player, so expectations of its performance are correspondingly high.
The Darblet is controlled via its own menu, which proves surprisingly versatile. Various application profiles are preprogrammed and can then be fine-tuned. As with the other systems, we chose a moderate setting.
In our first test scene, the Darblet also achieves increased sharpness at first glance. A closer look reveals how it works: in fine, high-frequency details such as hair, gamma is increased to raise contrast and thus subjective sharpness.
Top: original, bottom: Darbee Darblet
This method is also applied to coarser details such as the folds in the bag, so they gain in depth as well. The result looks appealing without being overdone. The houses in the background are sharpened slightly, but the original composition is not distorted too much. Distracting picture noise is not provoked either.
The algorithm deserves particular praise for its handling of the contrast transitions at the legs: they are neither sharpened, nor is the ringing in the original amplified. No processing takes place here at all – because none is needed.
Top: original, bottom: Darbee Darblet
Good results with St Mark’s Square too: detail is increased without giving the image a digital look. Small details keep their colour. Noise is amplified only minimally and remains inconspicuous. However, the subjective gain in sharpness is not quite as high as with Reality Creation.
3. Conclusion
Three different systems, three algorithms, three results: although each manufacturer promises – and delivers – the same thing, all three systems create an entirely different look that is laid over the image like a fingerprint, with individual strengths and weaknesses. In addition to the sample scenes above, we examined all systems with a wide range of material to gain a nuanced impression.
We found the Darbee Darblet’s picture the most balanced: once (moderately) configured, it noticeably increases sharpness while remaining subtle enough not to overemphasise distracting artefacts such as double contours, aliasing or noise. In most cases, the composition, including any intentional blur, is not distorted too much either. Despite its high price, the little image processor can visibly enhance the picture of projectors that are not equipped with dedicated sharpening algorithms from the factory.
Epson’s Super Resolution undoubtedly produces the quickest “wow” effect, as subjective sharpness is clearly boosted at first glance in almost every scene. This is pleasing at first, but on critical inspection various amplified artefacts quickly become apparent, such as edge flicker, noise, dithering and overexposure, as well as colour loss in small details. We recommend level “2”, which strikes the best balance between sharpness gain and fidelity. All in all, the Super Resolution algorithm is showing its age and needs a fundamental overhaul for future generations.
Sony Reality Creation finally sits in the middle: it produces less pronounced sharpness than Epson’s version, but brings out more detail than Darbee’s Darblet with 4K-mastered Blu-rays. However, its strong emphasis on picture noise can quickly introduce shortcomings that spoil the natural look. Here too, less is often more!
That leaves the question of how useful these systems are for accurate picture reproduction: well-mastered Blu-rays are undoubtedly sharp enough that you can do without additional “sharpening”. With mediocre transfers, however, which unfortunately still make up the vast majority of films available, they can effectively reduce shortcomings. We recommend a personal comparison at one of our home cinema partners across Germany.
31 January,
Ekkehart Schmitt
Cine4Home
Matching product categories in our shop:
Further guides:
- Projector buying guide using the BEAM method – overview of all guides
