File Formats Revisited

Digital photography has come a long way in the past decade. Cameras are offering higher resolution, better performance at high sensitivity settings and greatly reduced power consumption. More user-friendly functions (like live viewing) are being packed into ever smaller bodies, and prices are now affordable for even cash-strapped buyers. But file formats are one of the few features that have not changed.

However, change is well overdue and not technologically impossible to implement. What we need are some ways to change the firmware in cameras so point-and-shoot digicam users are no longer stuck with inefficient JPEGs and serious photographers don’t have to face a quagmire of conflicting and continuously-changing Raw file formats.

Promises, Promises
Sure, there have been promises of improvements. Eight years ago we thought JPEG 2000 (designated by the *.jpx extension) would offer some worthwhile advances in the form of superior compression performance, better resilience and support for both lossless and lossy compression. But that initiative was nipped in the bud by a total lack of support from any of the leading camera manufacturers.

Perhaps the possibility that undeclared and obscure intellectual property rights could affect implementation of the standard put some manufacturers off. However, the standard was supposed to be available without payment of royalty or licence fees. At the time, a few software developers included JPEG 2000 support and it’s still included in most Linux distributions so that can’t be the only explanation.

A version of JPEG 2000, Motion JPEG 2000, is currently the leading digital film standard supported by the Digital Cinema Initiatives consortium (which includes most major studios and distributors) for film distribution and exhibition. The first live-action stereoscopic feature encoded with JPEG 2000, Beowulf, was shown in cinemas around the world late last year. So why did still image capture miss out?

In mid-2006, Microsoft launched a proprietary still-image compression system, known as Windows Media Photo (*.wdp), which claimed to deliver compression quality comparable to JPEG 2000. The new format boasted a ‘lightweight, high-performance algorithm with a small memory footprint’ (in other words, it’s more efficient). It provides more than twice the quality of regular JPEG, which is a significant advantage.

To avoid the problems associated with proprietary formats, Windows Media Photo was renamed HD Photo (*.hdp) about a year later. Although Microsoft holds patents on the technology, it offers the specification for free and encourages open-source developers to make use of the format. Interestingly, it was not listed among the technologies covered by Microsoft’s Open Specification Promise when we went to press, so that could be one reason camera manufacturers are holding back.

In November, 2007, the Joint Photographic Experts Group (JPEG) proposed the adoption of HD Photo as a new compression standard under the name, JPEG XR (extended range). The JPEG XR specification (Working Draft) will be balloted for promotion to Committee Draft (CD) status before the 44th WG1 San Francisco meeting, March 31 to April 4, 2008.
Will the camera manufacturers adopt it? Who can say; but consumer demand – particularly from enthusiast and professional photographers – could exert some pressure upon them to do so. A new, improved ‘universal’ file format is definitely overdue.

What’s Wrong with JPEG?
The downside of JPEG is that it is always lossy. Compression is achieved by dividing the image into small blocks and discarding data that is unlikely to be missed by the human eye. This ‘lost’ data can never be recovered.

The degree of compression is, however, adjustable. Most digital cameras provide at least two compression levels in the image ‘Quality’ setting. Some offer more.

To complicate matters, each time you open an image for editing, when it’s re-saved the file will be compressed again. By the time you’ve saved the file eight or nine times it will be both smaller and lower in quality. Heavily compressed images can look blocky on computer screens and usually make poor quality prints. Excessive compression can also produce JPEG artefacts.

JPEG images are also 8-bit images, which means you have only 256 (2⁸) levels of red, green and blue per pixel to work with when editing them. It’s not a problem unless you want (or need) to edit the image files. Eight-bit images may not have enough data to recover shadow if you have underexposed shots. They also provide absolutely no flexibility for recovering blown-out highlights. The risk of murky shadows, increased noise and posterisation when editing even slightly flawed JPEGs is very high.

The above illustration shows a typical JPEG file in which parts of the image have little detail, while other parts have a great deal.

A maximum-size enlargement of the edge of the right hand subject’s hat shows the difference in structure between the hat (which contains some detail) and the almost uniform sky. Much higher levels of compression are used for the sky because more data can be discarded without viewers noticing.

The same-sized enlargement of the details on the left side subject’s hat shows the blocky structure created by JPEG compression.

Raw Files
To overcome all the problems associated with JPEG capture, most serious photographers shoot Raw files. Although they usually contain embedded JPEG images (which are used for displaying images on the camera’s LCD or in browser software as well as for histogram calculation), Raw files are essentially un-processed.

The data in a Raw file is simply a log of the number of photons each photosite has collected and converted into electrons. This analogue value is converted into digital form by the camera’s analogue-to-digital converter.

Some manufacturers tweak the data to bias white balance; others apply noise filters to deal with long exposures. But, all Raw files must be processed post-capture to produce editable images.

The main advantages of Raw files relate directly to post-capture processing, and one of the greatest is the ability to extract images with at least 16 times more data than JPEG images. The more data you have to work with when editing images, the more flexibility you have to make adjustments without running into problems like murky shadows, increased noise and posterisation. A 12-bit Raw file will give you 4096 (2¹²) levels of red, green and blue per pixel, while a 14-bit file provides 16,384 (2¹⁴) possible values per pixel.

Sounds great? It is – but wait, there’s more!

What’s Wrong with Raw?
Raw files are far from perfect – and they are nowhere near as easy to use as JPEGs, which is why few digicams offer Raw file capture. Here’s why:

1. Because they store so much more data, Raw files are much larger than JPEGs. Although some manufacturers compress their Raw files (usually losslessly), a typical Raw file from a 10-megapixel camera can range from about 9MB to 15MB in size. This is much larger than a high-resolution JPEG with the same pixel dimensions, which is typically between 3MB and 5MB in size. (With the cost of high-capacity memory cards plummeting, this is less of a problem than it once was.)

2. Raw files are proprietary. Each manufacturer has its own version and most manufacturers change their Raw file formats with each new generation of cameras. New Raw file formats are seldom compatible with older versions of the conversion software, so you need new converters each time you buy a new camera. Older cameras sometimes end up being unsupported by proprietary converters.

Adobe attempted to standardise the Raw file format with the release of its Digital Negative Specification (DNG) in September 2004. However, only a few major manufacturers (currently Hasselblad, Leica, Pentax, Ricoh and Samsung) offer DNG as a Raw file format in their cameras – and some only in selected models.

In its latest DSLR cameras, Pentax supports Raw file capture in both the proprietary PEF.RAW format and the open DNG.RAW.

3. It can take some time for manufacturers to update their Raw file conversion software to include new cameras. This issue applies to both the individual camera manufacturer updating its proprietary conversion software and third-party software developers like Adobe, Bibble, PhaseOne and others bring ing out updates that offer the full range of adjustments photographers require.

We’ve found occasions where the third-party developers have offered updates for new cameras a week or two before the manufacturer updated its proprietary software, which isn’t a good look. This problem wouldn’t happen if everyone supported DNG.RAW, even as a second-string format.

4. Some manufacturers (Nikon and Olympus being the leaders) charge extra for a fully-functional, proprietary Raw file converter for their cameras. Buyers of these brands receive a very basic browser application with the camera, leaving them to decide whether to invest additional dollars in the ‘full bottle’ converter or wait until one of the third-party developers jumps in to fill the breach. (Again, not a good look.)

5. Raw file converters are a moving target. Even proprietary converters vary hugely from one manufacturer to another – and usually from one generation to the next.

6. Some converters ignore part of the image data and may only apply some of the camera settings you have used when the picture was shot. It may also be difficult to access all of the Exif data when the Raw file is converted into a TIFF file for editing. And, even if a converter uses a particular setting, it may interpret the data differently from other converters. You have to keep trying out new converters to find out which one does the best with your image files and workflow. (Fortunately, most developers offer trial downloads.)

7. Most converters do more than simply convert Raw images into TIFF or JPEG format. Recent additions include adjustable sharpening, noise reduction processing, corrections for lens aberrations (vignetting, distortion and chromatic aberration) and highlight/shadow adjustment. These corrections usually involve simple post-processing but, in the process, they may introduce artefacts like softening, edge effects and posterisation.

8. Raw shooters must be particularly vigilant about colour space settings throughout their imaging workflow. Raw files have no inherent colour space and most of them will record a wider colour gamut than the traditional sRGB and Adobe RGB colour spaces. Even though you may have set a colour space on your camera, this is recorded as a tag only and can be changed post-capture. Applications like Photoshop Lightroom and Apple Aperture offer wider-gamut colour spaces like ProPhoto RGB, which you can use for editing. Larger colour spaces minimise the risk of posterisation and other data-related errors.

What can be done?
Solving these problems lies mainly with the camera manufacturers, because only they can offer the new and improved file formats. Will anything be done? We doubt it. To date, with the exception of the manufacturers listed above, proprietary Raw file formats continue to dominate – and proliferate.

Back in 2005, frustrated by on-going problems with proprietary Raw file formats, a group of photographers formed the OpenRAW initiative (www.OpenRAW.org/). Its aim was to encourage camera manufacturers to adopt an ‘open’ Raw file format that would be both universal and future-proof.

The group conducted an international online survey between 31 January and 15 March in 2006 to allow photographers to express their views on the issue. Of the more than 19,000 photographers who responded, over 90% felt ‘Camera makers should publish full and open descriptions of all parts of the Raw image files their cameras produce.’ And more than 66% were concerned about being able to open and edit Raw files created by older digital cameras.

Unfortunately, the OpenRAW website appears to have gone quiet in the past 12-18 months, leaving photographers without a place to express their frustration. The above problems continue to provide unnecessary hassles for photographers at all levels – and we STILL don’t have an ideal file format for image capture.