On November 3rd 2009 DxO released version 6 of DxO Optics Pro, a high-end RAW converter. So far, only the Windows version has been released. DxO Optics Pro, sometimes known simply as DxO or DOP, is best known for automatic corrections of lens errors – especially when converting a RAW file to a JPG. Its main advanced automated corrections are for lens softness, distortion, chromatic aberration, vignetting, sensor noise and dynamic range. In addition, it has the usual basic manual photo manipulation functions like cropping, straightening, exposure correction and color temperature correction.
DxO and lens corrections
DxO actually works on both JPEG and RAW files, but if you are interested in this level of image enhancements, you are probably using it as a RAW converter. The idea is simple: even the best lenses have a number of lens errors. Some, like
- chromatic aberration (aka color fringing)
can be systematically measured in a lab (and yes, DxO has a real R&D department where engineers measure the performance of specific lens/body combinations). Once you know these errors, you can automatically compensate for them in any image. Some errors like softness cannot be corrected entirely for very technical reasons. But it helps, and sharpness corrections with varying levels of sophistication are routinely carried out nowadays – not just with DxO, but in virtually all digital cameras and in most post-processing software. Other lens errors that DxO can handle relatively easily and cleanly are vignetting, distortion and chromatic aberration.
Note that these lens corrections depend on more than just the lens that you used: they depend on the zoom and aperture settings and to some degree on the focus setting. So to characterize a lens you need to measure it at a series of zoom/aperture/focus settings. In all, DxO says they take more than 1000 images for any specific lens-and-body combination.
Note that all of the above corrections can alternatively be applied with products like Adobe’s Photoshop, Adobe’s Lightroom or Apple’s Aperture. But there you will need to correct these manually: the software does know how the image was taken, but these products don’t use this information to automatically correct the image. DxO does the opposite: it uses the EXIF information stored in the image to create what it thinks the image would look like if the lens had been (more) ideal.
Note that, although this helps the output of a lens look like it was made with a more expensive lens, even the best lenses (especially at the wide-angle end) are far from ideal when sampled at the current image resolutions. You thus shouldn’t expect miracles (like in Hollywood’s idea of photo enhancement), but you can expect a result that is better and faster than what you can achieve with manual image tweaking.
DxO and camera body limitations
DxO does something similar for limitations of the sensor and the camera’s internal image processing:
- noise (depends on ISO and lighting levels)
- color (depends on the body only)
- contrast (depends on the body only)
The latter two are minor: presumably the camera manufacturer does its best to fix this already. But some tweaking can be done if you don’t agree with these settings: they are after all a matter of taste because many consumers don’t see their camera as a calibrated scientific measurement device, but as a way to create good-looking pictures. Curiously, DxO even allows you to simulate what an image taken with one camera body would look like if it had been taken with another camera body.
Noise, however, is another story. Until recently, cameras were hesitant to reduce raw image noise in-camera and left this to post-processing software. Again DxO, instead of only providing 2 or 3 sliders with which to reduce color (chroma) noise and exposure (luminance) noise, uses measurements and advanced algorithms. It is quite a challenge to reduce noise while minimizing softening of the meaningful details. DxO and some of its competitors are pretty good at this, and started doing high-end noise reduction long before the camera manufacturers decided that high-ISO was a marketing hype.
DxO is thus all about automatically enhancing images, and then giving the user the opportunity to provide the traditional set of common manipulations (cropping, exposure, adjusting the color temperature, etc). In DxO Optics Pro v6, DxO added the option to do some of the automatic corrections entirely manually as well. This is not because manual might be better than automatic, but simply because DxO Optics Pro v5 could only automatically correct certain errors for camera/body combinations that it had tested. If your lens/body combination is not one of the 1600 currently supported ones (the number keeps growing as new equipment is released and existing equipment is measured) you can now resort to adjusting manually like in Lightroom or Aperture.
Have Lightroom and Aperture adopted DxO’s automatic approach? Surprisingly so far, they have not. I am still expecting this to happen at some point. In fact, I always suspected that DxO might be bought by a larger company for their technical expertise and have their core functionality integrated into Aperture, Lightroom, Picasa, Canon DPP or whatever: these post-processing products have a much larger market share than DxO and thus presumably have larger budgets than specialist DxO. But again, this hasn’t happened and this thought is purely speculation on my part.
What has happened in the past year or so is that some of the DxO automatic image correction functions are migrating into the cameras, especially for JPG images. That makes a lot of sense if you are Canon or Nikon: you are very aware and competent with respect to lens errors and their impact. And can “teach” your cameras to make the necessary corrections using digital signal processing.
The most obvious example of “safe” in-camera image defect removal is vignetting. That can be characterized per lens at a given aperture with a simple table of correction values (exposure correction depends on distance from the image center, focal length and aperture). The image can be improved by multiplying each pixel with a correction factor. That may require say 10 million multiplications (early in the image pipeline) or (with fewer multiplications) by tweaking the JPEG output. I believe that some camera’s also correct chromatic aberration in-camera: shift the red, green and blue images to correct for lens faults. It is likely that this trend to correct camera limitations in the camera itself will slowly continue as camera manufacturers master the technology and as users grow to accept more intelligence in cameras.
User response to DxO Optics Pro v6
A look at the user forums at DxO (you need an account to get in) suggests that performance has improved, some user interface modifications confuse existing users (“where did it go?”), noise reduction at extreme ISO is improved, and a normal level of initial problems with any new major release. In contrast, the initial release of DxO Optics Pro v5 had quite a few stability issues around its user interface (the core processing algorithms of DxO always seem to have been quite dependable).
My experience so far
I have not done extensive testing (e.g. was not a Beta tester for v6 – although I was invited to become one) and have only used DxO Optics Pro Elite v6 after its release to get some specific jobs done. I use DxO next to Lightroom to apply enhancements to my best pictures to do things Lightroom cannot do (lens corrections, advanced lighting, keystoning). Although I paid for the upgrade and then installed it, you can test DxO for free for one month before activating it with a key.
In the finest DxO tradition, build 7502 of the new version 6 crashed during its first task. As a former heavy user of version 5 I immediately tried the heavy task I was needing to do: processing of a 21 MPixel (Canon 5DII) image exported from Lightroom to DxO via what DxO calls “Scenario 2”. This means that Lightroom generates a TIFF 16-bit uncompressed image, adds it to the database, and invokes DxO to update the TIFF. I originally thought this was related to the huge (120 MByte) TIFF file or to the Lightroom integration: after all, I know that the software had a multi-month beta testing program, and know that DxO wanted to avoid the problems they had when they released with DxO v5.0.
Incidentally bugs like this can be filed by users in a bug tracking database on the DxO site. I recommend doing this (although I admit I cut a corner by sending the log file to the DxO customer support department). A day or two later, they told me that they had tracked down “the root cause” of the bug (I probably wasn’t the first to see it) and had a workaround until a new release became available: in the “Process” tab, click “Show output Settings panel” before the “Start processing”.
User interface improvements
The user interface has been improved significantly. It had a better feel than DxO v5. One example: the colorful icons for classes of operations have been replaced by text. To me the icons were irritating because, even with extensive use, they were not that intuitive (maybe because the icons represented groups of operations rather than individual operations).
Another useful improvement is that the operations can be shown in beginner-, intermediate- and expert mode (they are called “first steps”, “essentials” and “advanced user”). This can serve to hide less frequently used manipulations. Complex operations (e.g. keystoning) also have the option of showing basics versus showing all available sliders. This worked fine for me.
Another improvement is the loading of lens modules. In DxO v5 this was less intuitive and didn’t work smoothly. Now, I downloaded the modules before running the rest of the software and this went cleanly and smoothly. And when I stumbled on an image taken with a lens I no longer own (Canon 75-300 mm IS), it offered to download the module.