Combining multiple poor quality CCTV images into a single, computer-enhanced composite could improve the accuracy of facial recognition systems used to identify criminal suspects, new research suggests.