PhD Studentship: Human Referenced Colour Processing

We live in an era where machine learning (ML) based systems can solve a wide variety of image-based problems. However, the architectures deployed are not biologically plausible and the training required to use them can be orders of magnitude more complex than that available to human vision. Further, there are still visual competences where we as human observers outperform even the most complex networks. One such competence is dynamic range compression (related to tone mapping) where an initial image has its dynamic range reduced. In dynamic range compression (DRC) a large 10,000:1 signal range e.g. images of scenes where there is strong sunlight and deep shadows - is compressed by two orders of magnitude to 100:1. The human vision systems and ML-based algorithms attempt to compress the dynamic range so that the bit-depth compressed image `looks good. DRC is a key component in all photographic processing pipelines.

In this project we will investigate the similarities and divergences between architectures for solving the dynamic range compression in human and machines. Our aim is to design an architecture that is consistent with human vision but where modern machine learning techniques might be deployed to optimise performance. Our work will build on Land and McCanns seminal Retinex theory [1], its interpretation in the context of known brain physiology [2] and its modern re-engineering as an effective numerical algorithm [3]. We will also investigate why the current non-human-vision-referenced ML algorithms produce poor results [4]. The research will take place in the Colour & Imaging Lab at the University of East Anglia.

[1] Land EH, McCann JJ. J Opt Soc Am. 1971;61(1):111.

[2] Hurlbert A. J Opt Soc Am A. 1986;3(10):16841693.

[3] Karami A, Finlayson G. London Imaging Meeting. 2024;5(1):1721.

[4] Karami A, Finlayson GD. AIC 2025. 2025; 143-148.

Entry Requirements

The standard minimum entry requirement is 2:1 in Mathematics, Computer Science, Physics and other numerate disciplines.

Mode of Study

Full-time

Start Date

1 October 2027

Additional Funding Information

This PhD project is in a competition for a Faculty of Science funded studentship. Funding is available to UK applicants and comprises home tuition fees and an annual stipend for 3 years. The current UEA / institutional PhD stipend rate is 21,805 and will be subject to any agreed inflationary uplift for 2027/28.

Advert information