Vision for a brighter future

Researchers at Oxford University are developing technologies that could improve the dynamic range and colour of images produced by digital cameras without large increases in their price.

Researchers at Oxford University are developing technologies that could improve the dynamic range and colour of images produced by digital cameras — without large increases in their price.

Compared to human vision, digital cameras have two critical limitations. The first is their dynamic range; scenes that have both bright and dark areas are said to have a high dynamic range and the dynamic range of some scenes is too large for existing cameras.

The second, usually less noticeable, limitation is the variability in object colour under different lighting conditions.

‘The problem is that digital cameras are very successful, but they are not as good as they should be, and in some lighting conditions you get saturation,’ said project leader Dr Steve Collins of the department of engineering. ‘For instance, if you try to take a picture of someone standing in the shade on a bright day, then he or she tends to disappear. The camera misses some detail the eye can see.’

This means that for most users, cameras are only really suitable for taking pictures under controlled light conditions. Serious amateurs can get round the problem by taking multiple shots of a subject using different exposures and a computer to stitch the images together. The problem is that this technique can only be used for a static scene, because if there is movement between shots, the multiple images can no longer be successfully blended.

Some camera manufacturers are attempting to automate this process, using sensors that can take a number of images quickly after a single press of the button. However, the amount of data that is then stored within the camera is very large. The technology is also unable to guarantee that the colour of the captured images will be correct, and the equipment is more expensive than a normal camera.

The Oxford team, consisting of Collins, together with Dr Paul Burn of the department of chemistry, is trying to increase the dynamic range of digital cameras while preserving their ability to capture colours accurately. This is achieved using the smallest number of bits of data possible as this is what will keep the cost of the parts needed to process the images to a minimum — allowing the technology to be accessible to the mass market.

The three-year project starts in December, and has received EPSRC funding of just under £500,000.

The technology will use pixels that respond to a narrower-than-usual range of wavelengths and create an output that is proportional to the logarithm of the amount of light.

Using these ‘logarithmic’ pixels will allow the camera to avoid light saturation while retaining the important information in the scene in fewer bits. The ultimate aim of the project is to produce pixels whose output can be processed so that the colour of an object can be reliably captured in different lighting conditions. These will be used to create a point-and-click digital camera that can produce a detailed image closer to that which can be seen by the eye — regardless of lighting variations.

As silicon does not possess the right spectral characteristics to achieve all this, the team will be searching for a different material to create organic photodiodes. The researchers will also have to design the electronics to gather the information from the new material.

The result will be a device that would be able to provide good dynamic range and colour reproduction at a low price. ‘At the moment, cameras are trying to differentiate themselves using software features,’ said Collins, citing the recently-released svelte mode as an example. Using this, the subject of the picture is stretched, making them look taller and thinner. ‘However, our work will enable cameras to be differentiated through easily achievable better picture quality.’

Previously, the driving force behind the camera industry has been focused on improving the devices’ megapixel count. However, there is a growing belief that the megapixel race has run its course, as seven and eight megapixel cameras are more than adequate for most applications.

The team already has patents pending on technologies that can provide the high dynamic range, and the narrow spectral response.