RAW Titans Face Off - II - Feature Comparison
by Thomas Theuerkorn, July 2007
RAW conversion software all have the same purpose: to develop the RAW "positives" and generate pictures that can be read by any program for viewing, printing or electronic publishing. Beyond that common denominator, the approach to improving picture quality can be quite different between the programs. It's easy to claim that one program is better than the other, but in the end it's achievable picture quality in certain situations that sets them apart. Some of it depends on the used equipment, others are simply a measure of the developer's ingenuity.
For a general overview of the contenders, please [Click Here] for a feature list and and an explanation what RAW actually means. However, the "proof is in the pudding" and in this overview, I am going to look into vital functionality to solve a number of problems, which in return is the basis for achievable quality. Of course the best results are still easiest to achieve the better the actual photo is. So pay attention to equipment, setup and exposure; but with that taken care of, the remaining bit of improvement (and in difficult situations) it becomes a matter of software and mastering the functionality it provides.
While there are other programs (like Silkypix or CaptureOne), I am going to concentrate on five that I am most familiar with. With exception of Bibble Pro, either tool can be found at some point in my typical conversion process. (Upon popular demand I included a 2-month evaluation version of Bibble Pro to round things up.) In this test, five programs (Canon Digital Photo Professional 3, DxO Optics Pro 4.5, Adobe Lightroom 1.1, LightCrafts Lightzone 3.0.6 and Bibble Pro 4.9.8b) are evaluated regarding major tasks to improve picture quality, like ...
- White Balance
- Chromatic Aberration
- Noise reduction
- Selection guide
[Exposure] - Given that photography is nothing more (or less) than capturing light to record what the eyes see in a particular moment, it doesn't matter so much how that light was created. What's important is that the end product matches your expectations and corrections are often necessary to achieve that goal. Lucky you, it has never been easier than with digital photography where light is represented by color and associated brightness values. Should you ever be so lucky that both light and camera settings turn out to be just perfect, you may not need correction. However, digital technology has its limits and many situations require more effort due to measurement difficulties or simply available exposure range. In fact, in some ways it's more restrictive since a blown out highlights are pretty much irrecoverable, where film technology typically has a bit more room until such limitations apply. With exposure representing a range from the darkest colors to the brightest, the histogram has become a valuable tool in determining the quality of the captured data. The tone curve is an easy way to manipulate that distribution by boosting values based on their brightness (or color). More sophisticated are advanced methods like fill light and highlight recovery, both geared to deal with either insufficient exposure or excessive dynamic range.
It should not be a surprise that LightZone pushes hard to take top honors in the exposure category. Zones or not, it provides the most control over data in a histogram by dividing the histogram range into 16 zones which can be individually controlled. The transition between two zones is also a handle to move that reference point towards darker or lighter, and the neighboring zones are equally affected. Automated RAW adjustments help getting to a balanced shot quicker. With all the adjustability comes complexity for beginners and sometimes a more distinct manual approach to a number of standard solutions. More traditional are Lightroom and Optics Pro which both deploy the traditional tone curve and "automated" distribution changes to improve overall contrast. DxO lighting in general is a bit aggressive and may not be the best option as default, but the fine adjustments allow for powerful distribution manipulation and extract detail from virtually any situation while largely maintaining each extreme of the spectrum. However, be advised that it doesn't distinguish between intentionally and accidentally over- or underexposed shots and does require supervision. Canon's Digital Photo Professional 3 offers the most basic functionality with a relatively restrictive tone curve and no fill light or highlight recovery. At least the white and black point can be adjusted to maximize the output dynamic range. Clarity or localized contrast have slightly different implementations between the programs, but only DPP3 doesn't offer any similar functionality and Bibble's "Perfectly Clear" is slightly unpredictable.
A simple daylight scene at Colonial Williamsburg (at noon)
[White Balance] - This is strictly speaking not an exact setting, but more of a science with a lot of variables. Yet, the user doesn't have to write a dissertation to correct his or her pictures. To make things easier, typical correction consists of two parameters: Temperature and Tint. Temperature is measured in Kelvin and expresses the needed temperature of a Black Body to be heated to in order to emit light of the equivalent spectrum. Tint, on the other hand, adjusts for imperfections in that spectrum due to the inability of a certain light source (i.e. fluorescent) to recreate the full spectrum associated with determined temperature. It doesn't make it any easier, that the human brain adjusts for different light sources by shifting our perception of White accordingly (since we "know" that a piece of paper is White). Take an incandescent light bulb for example, a low power design would glow at a lower temperature (typ. 2600K) than a high power bulb (typ. 3000K). Yet, our brain uses actual fire as a reference for "hot" (despite it's even lower temperature of roughly 1600K) and we are more likely to describe the low power bulb's light as "warmer" than the high power bulb. Daylight at noon (avg. 5500K) is much "hotter" but neutral to our perception (thanks to mother Nature and the fact that the sun is much further away and feels "cooler" than a fire up close).
It get's a bit easier in the digital world as any shade of White (incl. Gray and Black) is neutral which means they consist of "perfectly" balanced color components for Red, Green and Blue (i.e. R50% + G50% + B50% = Gray50%). Hence, if there is such a known neutral reference in the picture, the computer can easily figure out the deviation in the original picture and adjust accordingly. That's the background behind so-called Gray cards which provide that known neutral color. (A full White would be in danger of being washed out and unreliable source for color temperature.) If that sounds too theoretical, all one really needs to know is that Temperature can range from cold to warm (blue to yellow) and Tint either adds Green or Magenta to compensate for deviations. Goal of the compensation is to bring the color spectrum back to our reference (Daylight) for "true" colors, where desired.
White and daylight are apparently open for interpretation. (To compare "As shot" with "Daylight" roll over each picture.)
However, unless you're using a calibrated Gray card like WhiBal, it's sometimes tricky to guess the correct lighting. It's easy to say it's sunny or daylight, but even that's by no means the exact same light depending on time of the day, season and location. It's helpful to have a known neutral colored object in the picture to color pick the white balance. Nevertheless, presets can help out where no such information is available. Most applications have such settings included to allow for quick results. Yet, each implementation is slightly different and automated results more or less just a matter of convenience.
Lighting temperature by different methods (tint not shown)
It's quite surprising to see how differently each program interprets "As Shot" white balance. The scene was shot at normal daylight and with Auto-WB setting on my EOS 20D (not ideal, technically the same input for every program). One could assume 5500K and no Tint to be the right setting, yet Lightroom claimed 5200K, Lightzone 5922K, and Bibble Pro even 4727K. The latter was heavily compensated via magenta tint which brought the overall appearance closer to normal than the number suggests. Not surprisingly, DPP3 returned with 5600K the closest result, but it's also the one with possible insider knowledge about the sensor. DxO Optics Pro seemingly nailed it with a perfect 5500K and only very slight magenta tint, but the picture still has a slight yellowish tint.
When forcing "Daylight" white balance, one would expect for the difference to vanish and every program to set the temperature to an agreed upon standard 5500K (without any tint). However, differences were quite obvious. Most surprising is Optics Pro with dropping color temperature to 5200K and overall even more yellowish appearance. Bibble recovered and got closer to a balanced shot while still maintaining a slightly warm rendering. Surprisingly the temperature now even reads only 4500K. Lightroom switched places with Optics Pro and the "As Shot" (5250K) now turned into a 5500K rendering plus more magenta tint (which became obvious). It's a bit harder to verify DPP3's settings for daylight since it's not disclosing the number but Canon typically defines 5200K as daylight. LightZone, on the other hand, doesn't provide any presets and relies on the user to move the temperature setting to the correct white balance. This may be a disadvantage for quick adjustment, but overall neither program can guarantee perfect results with canned settings anyway.
This is where the color picker for white balance is highly recommended for any program, along with a neutral reference like WhiBal Gray Card or the Macbeth Color Checker. What happens here is that the user defines a spot in the picture to consider neutral and the algorithm hunts along the Blue-Yellow line (Planckian locus) for temperature to find the best fit. Remaining discrepancies are then attempted to be minimized by searching along the Green-Magenta tint line. As algorithms differ, so do the results but it's quite visible that "As shot" and "Picked" are much closer in actual color temperature than the canned "Daylight" setting for any program just closer to the goal of removing discoloration. The big numerical deviation in Bibble is only to explain with algorithm differences which may favor tint correction over temperature, but the temp reading is certainly deceiving and around 4500K is too low for the lighting condition. (According to established references, daylight ranges from 5000K to 7000K). It's even more so surprising that the strong tint manages to compensate for most of it.
Last but not least, two programs allow something special. Lightroom and Lightzone both make it possible to treat shadow areas different in color tone than the brightly lit portion. Since Shadows are generally roughly 6500K and with normal Daylight referenced at 5500K, "split toning" allows Lightroom to bring both closer for a more balanced picture, but settings act global and don't discriminate between true shadow areas and similar tones in the brightly lit portion. Lightroom, on the other hand, allows to theoretically correct an unlimited number of regions to be balanced for their own color temperature. So while overall accuracy of both Lightroom and Lightzone is behind DPP3, this feature explains the higher rating of both.
McQuiggan (our cat) puts noise reduction to the test.
[Noise Reduction] - As the available light drops, increased sensor sensitivity is required and the associated amplification also raises the artifacts that are known as noise. This being a law of physics (for electronics), it's only a matter of how pronounced the noise is with any given design. Don't be fooled by the JPEG images coming out of your camera. Those have typically in-camera noise reduction applied. However, this may also reduce sharpness. To users that are new to dSLRs, it may come as a "shock" how much noise can actually be in the RAW file. Hence, some come to the premature conclusion that JPG is "better". Yet, RAW gives the user the opportunity to tweak noise reduction and apply different algorithms for best results.
While the best recipe for low noise is still proper lighting and the slowest ISO speed, fast moving subjects may not allow for such comfort and shadow areas can even in properly lit setups pose a challenge for the camera. Simple noise reduction blurs the picture without discriminating the details and ultimately reduces picture quality almost as much as the noise caused grief to begin with. More intelligent algorithms try to strike for a balance between noise removal and detail retention. In this picture, the eye poses such problem where maintaining the crispness of the iris is crucial for the composition.
Below is a comparison of each method's efficiency in tackling noise while maintaining detail. It should be noted that noise can vary between situations, and results may not be comparable to the full extend. For illustration purposes, look at the details below and how each program managed the noise in the right eye. (Hover over tab to see result.)
Lightroom takes the lead in noise reduction. Somewhat disappointing is Optics Pro with remaining color artifacts..
The results point to Lightroom 1.1 as the winner. It shows the best noise reduction with minimal effort and virtually no loss in detail (sharpness). Surprisingly good is the output from DPP which leaves only a bit luminance noise for maximum retention of detail, but most likely a result of the high applied RAW sharpening. This may be another home advantage for Canon, but it's impressive nevertheless. Lightzone offers RAW color noise reduction, which is very effective for chrominance. Optics Pro, on the other hand, requires a bit of tweaking despite not offering a whole lot of control and essentially cannot eliminate all the color artifacts to extract all detail. Bibble Pro uses NoiseNinja and excessive smoothing is needed to get rid of the noise but also looses the most detail in the test. The full registered version may return better results than the standard version of Noise Ninja that's included. The full "Noise Ninja" isn't available until you register for a full license. However, it's unlikely to dethrone Lightroom.
[Sharpening] - Putting the sharpening to the test with a detail of a church near my birth place. The 300 mm lens rendered good detail, but slightly soft. Hence the software needs to take it further. What should you look for? Well, since the original is an 8 MPx picture, it's impractical to give you the whole picture in all six variations. The four selected regions show specific problems ...
- Most of the roof detail needs to be smooth and not banded. Is the sharpening selective or global? The sky needs to be smooth and not grainy. The crown of the roof detail should be crisp and free of halos.
- Look at the pigeons. How much detail can the respective software extract? Further look at the arch of the window and how sharp the edge is rendered.
- The main feature here is the dark edge between the shadow and the sky shining through. High contrast edges like that are typically offset with a bright line to create the illusion of sharpness. However, nowadays this is more or less a sign of weakness since the halo effect pretty much screams "over sharpened".
- The mesh in the window (bottom) is a one-pixel feature which can easily disappear or bloom to be overly emphasized. Look for the program that makes the mesh pop and remain visible. Further, look at the lightning rod to the right and how crisp it's rendered without the halo effect.
Now that you're oriented, let's look into each program. The only parameters I played with were sharpening and local contrast (where available). I tried to achieve the maximum sharpness to the point where problems just about become visible or when extreme contrast simply caused excessive banding. (Hover over the respective tab to see how each program performs.)
It can be seen that -- aside from the fact that each interprets "As shot" white balance differently -- the top three are a close battle with the top honors going to Lightroom for an amazingly intuitive masking and the mild sharpening induced halo effect. However, it's either best sharpness or no halos as especially detail (1) shows. Local contrast control, and adaptive masking is limited and prevents from the maximum score. Overall some features remain slightly soft (i.e. mesh in #4) with slight halos appearing. DPP3 could easily take the first place if there were local contrast controls and the ability to tweak sharpening with a mask. However, it demonstrates that RAW sharpening and "insider knowledge" can be very effective and DPP3 returns the crispest details in this test (with slight halo penalties). Just look at the pigeons (2) and roof top (1). Overall the remaining ranking is hard to determine, as Optics Pro offers incredible control over local contrast and lighting which in itself make the picture appear crisp(er). The sharpening, however, is prone to show halos when overdone but is already at normal settings quite effective. It's selective and the sky never becomes grainy. This would be an argument against Lightzone, as the process is not discriminating and relies on the manual regions (not done in this test). Details (1) and (3) show easily recognizable grain in the sky even without reaching the sharpening level of both Lightroom or DxO. Bibble Pro, seems a bit handicapped as halos and grain limit the amount of sharpening applied. This effect is much worse when "Perfectly Clear" is applied and virtually every edge becomes a source for halos. (not shown)
Note: The Relight tool in LightZone 3 now includes a "Detail" adjustment which pretty much is a local contrast tool and improves sharpness similar to Optics Pro. However, it's not easy to isolate this effect for a fair comparison within this scope and therefore the
[Chromatic Aberration] - This artifact is introduced by the lens design, and most pronounced in wide-angle objectives and zoom lenses. Even the most expensive lens design is potentially to show some aberration in high contrast areas (especially strong white features with very defined sharp edges). The effect is caused by slight differences in focal length depending on the wavelength of the light. "Perfect" White light holds all wavelengths of visible light, and a lens is typically design after the sensitivity of the human eye (at the midpoint or Green). The same refraction index (as found in a lens) will then treat each wavelength slightly differently and a thin line would spread into a microscopic rainbow. (I am sure everybody remembers the experiment with the prism at school.) This effect is then called chromatic aberration since essentially color separate and form color fringes at contrasted transition (where the bleeding is noticeable).
Neither Digital Photo Pro 3 nor Lightzone 3 correct for chromatic aberration. Of the others, only Optics Pro 4 is automated.
In this test only Digital Photo Pro and Lightzone don't provide correction for this lens flaw. However, it's noticeable that Canon apparently has a mild reduction built in while Lightzone does not. Lightroom and Bibble are very similar in the sense that they both rely on manual adjustment to compensate for the color bleed. All Optics Pro 4 requires is a single click (if your camera / lens combination is supported). This can be a great time saver. Results are comparable in quality to the best settings in Bibble 4 and Lightroom 1. Of the two latter, Lightroom is the easiest to adjust due to a more responsive interface. It should be noted that the color saturation in the samples above has been greatly boosted to highlight the effect, and in reality a good lens often provides less visible color fringes especially in the center. (All samples are actual pixels!)
Note: This should not to be confused with the false color artifacts that can be introduced during the demosaicing process. Chromatic aberration is an optical problem (lens) while demosaic artifacts are a sensor problem (math).
[Geometry] - Purists are likely to object and rightfully stress the importance of maintaining the original pixels for best sharpness. Geometry corrections in the software basically move pixels to a new location which essentially requires to interpolate the information in order to fill in missing data (for stretched sections) or omit data (for compressed section). While that may be a good argument against it, the only way around it are proper equipment and setup. Nevertheless, hand held shots are virtually impossible to level perfectly every time and your position dictates the perspective which often cannot be freely chosen (and not everybody owns the proper shift lens). Last but not least, especially a wide angle lens introduces spherical distortion which shows as bent lines. When the problem cannot be resolved with the optics, software needs to step up and save the picture. However, the needed tools are not as common as one might expect. I thought about a proper picture for this test and it just happened that a friend of mine suggested a brick wall for a similar purpose. This may not be the most interesting picture to look at, but geometry deviations are very easy to recognize and correction efficiency easy to judge. (Hover over tab to see results.)
Sophisticated geometry correction is rare in a RAW converter. Only one tool offers functionality to correct all three typical problems.
The picture intentionally contains three major geometry problems: (1) tilted horizon, (2) keystoning and (3) spherical distortion. Canon's Digital Photo Pro 3 does not correct any of those and leaves the problem to the optics and setup. Both Lightroom and Lightzone are equally restrictive, but they at least provide very useable horizon leveling which allows to draw a line to follow any given geometry to make horizontal. Bibble, on the other hand, provides manual entry of an angle in the respective panel and the user has to search for the "Straightening Cursor" icon to perform an advanced graphical horizon leveling (with grid display). Keystoning is available as optional plug-in. In the end, only Optics Pro 4 provides all three tools out of the box and the keystoning correction is especially powerful. In this case the Force Rectangle function addressed both horizontal and vertical keystoning in one step! Lens correction requires a supported module for your given lens / camera combination, but is less dramatic for medium to long lenses (unlike the used 17 mm focal length in this test). The true surprise is the retention of sharpness in Optics Pro despite the significant interpolation to compensate for the deformation.
[Demosaicing] - While all the features discussed so far more or less have their origin in the optical design of a camera, demosaicing is a problem introduced by the sensor design. With the Foveon sensor being the only exception, most matrix designs are derived from the cost-effective Bayer configuration. If you recall the principle behind the demosaicing process, a pixel is typically formed out of four sensor pixel (two green, one red and one blue). This is all fine if those cells receive the same information, but get's tricky when a transition happens to fall within a pixel. Say the bright side falls on one part of the matrix and the dark side on the other. Strictly speaking, the conversion would have to average between the two, but that is easier said than done since the converter cannot know which cell skews the input. The result is a false color generated by this differential spanning different color sensor cell.
Canon knows their sensors best, but Lightroom and Optics Pro also yield excellent results. Only Lightzone and Bibble have noticeable color artifacts.
Like in the chromatic aberration test, color saturation has been boosted in these samples for illustration purposes. While generally no converter can always guess what the missing information would have to be, it's quite impressive to see that Canon's expertise in both the sensor and the conversion software come to fruition with virtually no demosaic artifacts. Apparently the determination of an output pixel takes more into account than the directly contributing four sensor cells. Similar results can be seen in Optics Pro 4 and Adobe Lightroom, though with very faint tint spots in the white stripes (mostly Lightroom) which would be nearly invisible at normal saturation levels. Thanks to Lightzone and Bibble I can actually demonstrate the effect of false color artifacts which are more pronounced in either converter. Again, it should be noted that under normal saturation levels most of it would be very faint if at all visible. (All samples are actual pixels!)
[Performance] - Last but not least, let's look into the lofty speed claims each of the programs typically boasts in any press release. Each system will perform differently, and the following data is for comparison purpose only. With identical hardware, the results are fairly comparable. In this case an AMD Opteron 185 (virtually identical to the FX-60) processor has access to 2 GByte of DDR400 (C2) memory. The hard drive was a speedy Western Digital Raptor (10k rpm, SATA). The comparison is based on RAW files directly from a Canon EOS 20D (8.2 MPx). For the single conversion, the very same picture was used. For the batch process four similar RAW files were used. To the extend possible, each converter used all settings for a similar outcome but different feature sets naturally skew results. Lens and geometry correction was enabled where available. On the other hand, the Demosaic number reflects the attempt to measure a pure conversion without any options like noise reduction, sharpening, lens correction and exposure control. This is a rather theoretical number as the user is unlikely to perform a straight conversion. See how it compares to the more practical benchmarks to show how such features influence the overall speed.
Elapsed time [sec] for start-up, single picture and batch conversion (8.2 MPx RAW).
Canon's Digital Photo Professional 3.0.2 takes the undisputed lead with blazing startup speed and time to convert a single picture. The strictly sequential nature of the batch process is visible in the proportional conversion time. However, Canon "cheats" overall by holding back on advanced features. However, sharpening and noise reduction claim virtually no additional time.
Adobe's Lightroom 1.1 is a close second and even beats DPP3 in the batch process, indicating better use of dual core processor. Adobe profits from both a proven RAW engine and minimal geometry correction algorithms. Further, it's the fastest Demosaic process in the field, even beating DPP3 in the bare essentials.
DxO Optics Pro 4.5 shows a decent start-up and very efficient batch process. It looses some ground due to a relatively slow conversion of a single picture. However, it also deploys the most complex functions of the bunch. Nevertheless, even turning every function off to the degree possible (Demosaic only) shows that DxO doesn't have the fastest base conversion either. (GPU support in v4.5 doesn't currently work on my ATI hd2900xt.)
Bibble Pro 4.9.8 has a bit of a Jekyll and Hyde personality. Slow startup and great variability based on activated functionality are countered by fast basic conversion and the ability to reduce the pain by processing batches in the background while working on the next pictures. Enabling advanced features like lens correction, noise reduction and sharpening shows great impact on conversion speed. When advanced features are not enabled (Demosaic only), Bibble manages to nearly rival Lightroom's conversion speed and take second place. However, it's up to the user to decide how useful a straight conversion is. Unlike the other tested programs Bibble Pro is also more dependent on the hardware than expected. Another installation on an old(er) 3.0 GHz Pentium 4 with 2 GByte of RAM and 7200 rpm hard drive repeatedly clocked in at a whopping 145 seconds for the start-up alone. Oddly enough, it measured only 40% slower for the conversion (compared to the AMD Opteron 185). This system is not included in the benchmark above.
Lightzone 3 manages to edge out DxO in the conversion of a single picture, but only by a hair's breadth. The advantage get's a bit more obvious when reducing functionality to a plain Demosaic process. However, adding more complex operations (additional layers) easily looses that edge. Beaten in start-up and batch conversion, it takes last place when performance is the main decision point. However, the ability of combining an "unlimited" number of commands and regions does not only provide unmatched flexibility but also varies the conversion speed greatly.
I also included CaptureOne Pro 3.7.7 and RAWshooter Essential 2006 for reference. Both are a bit dated, but often considered to be very fast performers. Pixmatec's RAWshooter was actually purchased by Adobe at the end of 2006 and ultimately absorbed (into Lightroom?). PhaseOne's Capture One Pro 3.7 has been dormant for almost a year (and v4 is expected to release in 2007). While fast, neither is able to dethrone Lightroom or DPP3 in overall conversion performance. However, by looking at on-screen rendering it became clear that there is another side to speed. "Display 100%" describes the rendering time from Fit to window to displaying a picture at 100% (1:1) on the screen with no function activated other than demosaic. The "Preview 100%" is the same procedure but with several functions like sharpening, noise reduction, and lens distortion enabled (where available). In either case, both programs show good speed which translates into improved workflow. However, of all programs only Lightzone 3 shows excessive render time with mentioned features enabled for the preview. Overall, it's important to distinguish between display and conversion speed as both affect the workflow in different ways. (Note: Bibble Pro shows virtually instant performance which indicates that it actually works off a RGB representation rather than the RAW file.)
Note: This test does not take into consideration that one program might be faster on an Intel cpu (floating point) while another might gain more from AMD's processors (memory bandwidth) for comparable processor classes. Further, the applied algorithms can be very different in complexity and a fast program is only half the truth if the desired functionality is not available.
[Expandability] - Different users have different needs and a program can either try to serve them all or allow other to extend the base functionality with so-called plug-ins. While most programs can be configured as external editors in another host application, Bibble Pro 4 is the only one that also offers an open architecture for third party plug-ins. In fact, a few are included with the program like the B/W conversion which otherwise would absent in Bibble. Most useful extensions, however, are not free and with roughly $20 on average can turn Bibble into a rather expensive solution. Yet, this also keeps all the functionality in a familiar interface. This way, the user can actually add exotic features like the Zone system which otherwise is only available in Lightzone.
Lightroom is the only other choice to allow an import/export plug-in and use an external editor to expand its functionality. In this case both Dxo Optics Pro 4.5 and Lightroom 3.0.6 support to be such external editors. Before Optics Pro 4.5, the choice was clear and Lightzone 3 made a wonderful companion for Lightroom, but without ever solving Lightroom's lack of geometric correction. Now, Optics Pro 4.5 supports to be a LR plug-in and provides valuable optical correction to Adobe's program. For some users this may outweigh the improvements resulting from the Zone system in Lightzone. Further, Optics Pro also allows easy implementation of DxO's FilmPack, a solution to recreate the "charm" of certain film and development styles in the digital age.
Only Canon's DPP3 is neither plug-in nor host. Canon's own printing extension and custom color profiles being the only exception to the otherwise closed program.
[Selection] - In the end all the theory may not matter that much if the price or philosophy don't match your needs. Fortunately all programs offer downloads to test the software for up to one month and that should give you a good idea if it's for you. From this comparison it's clear that no single program does it all and well enough to make it the sole choice. So it comes down to your needs and application. Maybe the following recommendations might help to narrow down your choices. At this point I exclude DPP3 from those choices, despite its best-in-class sharpness and good noise reduction as well as blazing speed. Sure, it's free but also available only for owners of a Canon dSLR. However, if that's what you're shooting DPP certainly offers low cost and a powerful engine, but also requires a bit higher input quality.
Anyway, without any further ado, here the categories and the respective ratings (for relative comparison only) ...
|Bibble Pro 4||DPP 3.0.2||Optics Pro 4.5||Lightroom 1.1||Lightzone 3|
Note: The rating above does not take available plug-ins or other modifiers into consideration (except in the "Expandability)!