Every now and again, a product crops up that really grabs my attention. Recently, I noticed a short review of a product called FocusFixer. Despite the review being a bit short on detail, it roused my curiosity, and I investigated.

First impressions

FocusFixer is a product from the British company Fixer Labs. It makes a fairly incredible claim - to quote the product's ReadMe "FocusFixer is a Photoshop plug-in for Mac (OS X) and PC/Windows that restores out-of-focus images". This is a pretty astonishing claim. Sure, digital photos can - and need to be - sharpened, but all sharpening products, and pretty much every sharpening expert, states that sharpening can increase a photograph's accutance, thus increasing the sensation of sharpness, but an out of focus shot is out of focus - end of story. It seemed too good to be true, but since I have a vast collection of badly focussed photos, it was certainly worth trying. In particular, I have a shot from Jokulsarlon ice lagoon, in Iceland, which I particularly like, but which is badly focussed. So I downloaded the trial version of FocusFixer, and gave it a whirl. Approximately 20 seconds later my jaw hit the floor. Original image. Pretty, but as you can see from the detail view below, the foreground is soft. After running it through FocusFixer, a considerable improvement can be seen: First, it should be noted that I applied FocusFixer using a mask, selecting just the foreground ice above the waterline. Second, it is clear that the final result, whilst greatly improved, is still not going to win any competitions. However, it is now printable, and on an A4 print the difference is significant. It would have been better if I had focussed better at the time - but to err is human, and FocusFixer can help to reduce the pain.

The Product

So how does it all work ? Well Fixer Labs maintain a fairly inscrutable front, not giving too much away. They hint at a mathematical process which can refocus an image, perhaps somewhat akin to DXO Labs technologies. How all this works in practice I have no idea - clearly there is not enough information in a photograph to refocus it, in the way that coherent radar images can be focussed. All we have for each pixel is amplitude, no phase, no timing. But really, I don't care how they do it, just how well it is done. The software is implemented as a Photoshop plug-in, which provides a couple of simple controls. At the top are two before and after views, and a zoom control. Below these, two sliders. According to the user notes, Deblur gives a numerical feedback of the radius of the "circle of confusion" in pixels. The greater the effect you need, the higher you need to set the slider. I've found that a value between 4 and 5 is usually optimal. If you go too far, things get a bit wild. Threshold allows you to reduce noise and edge artefacts. I've found that it is usually better to keep Threshold at zero, and contain edge artefacts by carefully masking the area you want to work on. The next bit is intriguing: LensFIT (Lens File Information Technology) apparently is an optical modeling technology which uses camera EXIF data to identify the lens, and accordingly optimise processing. How it does this is not discussed, but there is certainly a subjective difference - an improvement - when LensFIT is turned on. For some cameras it will activate automatically. In other cases - including the Olympus E-1 I use - you have to give it a hint. FocusFixer seems to support a wide range of DSLR and digicam models, and more are being added. If the camera is not supported, a default algorithm is used. Now this could all be mumbo-jumbo, and I'm a bit puzzled as to how any optical modelling can be done without the lens information as well as the camera model. Certainly some information on the lens is in EXIF, but first I'm not sure that it is always adequate to uniquely identify a lens, and secondly it seems a bit unlikely that Fixer Labs has tested each and every lens on the market. DXO certainly haven't. However, it does appear to work, and the evidence is that there is indeed a new approach to sharpening underlying the plug-in. The fact that LensFIT has a patent pending doubtless makes it difficult for too much information to be revealed.

Field Test

FocusFixer is designed to correct focus blur. It cannot handle motion blur, and works best with high quality data. Focal Labs do not claim to work miracles, but they do deliver results. I decided to try out FocalFixer on a deliberately out of focus photo (not that I need to try hard) and compare the results with a similar in-focus shot. Because I'm lazy the shots, of a palm in sunlight, were handheld, at 1/200th - this may not be ideal. They were taken with an Olympus E-1 using the 14-54mm lens. The two images above are 100% detail zooms on a palm frond. The right-hand image is the "in focus" shot (unsharpened). The left-hand image has been partially processed by FocusFixer - the area of the palm above the red line is "fixed", the area below is untouched. I used settings of 4.5 Deblur, 0 Threshold. The conclusion is obvious: it is better to focus better! However, FocusFixer does a pretty good job of patching things up. The obvious question is can FocusFixer do things that cannot be done with Photoshop, or with other tools ? My answer is a qualified "yes" - qualified because I'm no Photoshop guru, and because I don't know all the tools on the market. Certainly I could not reproduce FocusFixer's results using Unsharp Mask (USM). With USM it was much harder to control detail, and edge artefacts and haloes become a real problem. The closest tool is perhaps the Creative Sharpener component of PhotoKit. This has a similar effect to FocusFixer, but is not so good at pulling out detail - at least not in my hands. On the other hand, it is suggested that FocusFixer used at very low Deblur settings might make a useful capture sharpening tool, but so far I see no reason to stop using PhotoKit for this task.

Downsides

There is room for improvement in FocusFixer. First, the preview is too small, or should at least be resizable. Secondly, the quality of the preview seems less good than the applied filter. Third, the product could do with a nicely written user manual - the ReadMe is a bit skimpy for a product of this price (although I'm not saying it is overpriced). Finally, the registration process is a pain in the neck. I'm not against companies protecting their rights, but in my case at least the process was needlessly complex and lengthy. Apparently Fixer Labs were suffering badly from piracy, and were compelled to protect themselves in this way. I'm still amazed by the all too common attitude that "copying" software is not theft - even amongst some software professionals. I do commend the license which allows non-simultaneous deployment on several computers owned by the customer, which is becoming a growing practice.

Conclusion

What Fixer Labs have tried to do is to bring to market a tool designed to do exactly what sharpening tool vendors claim cannot be done - fix out of focus pictures. The current version seems to do a pretty good job, and although I would not use it to replace other sharpening products, as a new tool in my digital workflow it is most welcome. FocusFixer costs $57 - for the same price, you can buy FixerBundle, which includes three other plugins, NoiseFixer, ShadowFixer, and TrueBlur. Fixer Labs also have released a resizing plug-in, Size Fixer, which I'm looking forward to trying when the Mac version is released. More information from the Fixer Labs web site.

Ok, I haven't written anything in this so-called blog for ages, but actually I've got a few things in the queue. The first is a brief write-up on using the E-1 for infra-red (IR) photography. Why IR ? Well it certainly gives a different perspective on things. Infrared photography uses either IR sensitive film, or an IR-sensitive digital sensor, to record an image at light frequencies lower than the human eye can detect, and to present a rendition of the scene which we can see. I have the distinct impression that whilst IR film is specifically designed to do this, digital camera sensors more or less do it despite the design. Most, if not all, digital cameras include an IR filter, I believe to reduce noise. This means that some are totally insensitive to IR, and those that are not require long exposure times, essentially to allow the light to leak past the filter. However, when IR photography does work with a digital camera, it has several advantages over film. First, IR film is very grainy (although some would not see that as a disadvantage). But the second is significant - you can check the exposure immediately after capture. This is significant, because exposure of IR film is largely guesswork and experience. There are no IR-sensitive lightmeters that I know of. A third advantage is the fact that you don't have all the problems associated with loading and developing IR film. This image is a photo taken with Kodak High Speed Infra Red film, using a Hasselblad Xpan. The standard characteristics of IR photography can be seen - the bright vegetation, the other-world effect, and in the case of film, the very coarse grain. Note that this photo was taken handheld, probably at around 1/60th second at f8 or thereabouts. The same photo with the E-1 would take somewhat longer. A number of different IR filters are available on the market. Some allow a small amount of visible light through, giving a deep, dark red effect. But "real" IR filters are blocking filters, so called because they let no visible light through at all. They appear black to the human eye. I tried two IR filters on the E-1, a B&W 093, and a Hoya R72. The B&W was too dark, but the Hoya works well provided long exposures are used. It goes without saying that IR photography with the E-1 requires a tripod, for two reasons - one the long exposure, and two, you have to frame the shot first with the filter off. Some digicams, such as the Nikon Coolpix 950 (still highly regarded by the IR community), allow you to preview the image on the screen. After a bit of trial and error, I discovered that on a reasonably sunny day - and you need sunlight for IR photography - using the Hoya filter requires exposures of between 45 and 60 seconds, at apertures between f4.5 and f11. I also quickly discovered that at such exposures the E-1's sensor shows a lot of thermal noise. Fortunately, switching on Noise Reduction fixes this, although at the cost of doubling the exposure time, which really taxes the battery. The two images above are 100% detail from a scene shot without noise reduction (left) and with (right). The image you get, after all this, is very red. Since the effect we're after here is to duplicate black & white IR film (colour IR film is another story altogether), we need to convert this to black and white. First of although you need to convert it from RAW (I don't recommend shooting JPG here). Don't use C1 for IR shots - obviously its well document difficulties with reds have a field day here. Use Photoshop ACR or Viewer/Studio, and ideally just go at default settings. If you use Photoshop ACR, you might see that the red channel is way overblown. Don't worry, but pull it back a bit using Exposure. Next, converting to black & white: Photoshop has several tools for converting to black and white, the most obvious being Desaturate. Don't use this. Use the Channel Mixer, preferably as an adjustment layer. The actual values you dial in to the Channel Mixer are a matter of taste, but generally most useful information is in the Red and Green channels - Blue is pretty noisy and dark. The image below was taken with the E-1 is bright afternoon sunlight, with an exposure of 60 seconds at f/11. In this case arguably it was slightly over-exposed. It was processed using the Channel Mixer settings shown here. So, in conclusion, IR photography with the E-1 works and is good fun. Long exposure times mean that you will need a still day, or will accept (or welcome) wind motion blur effects. It is different to film IR, but well worth exploring it its own right.