Content-Based Preferred Image Processing (FINLAYSONGU18SCI50)
Prof G Finlayson
Dr M Mackiewicz
No more applications being accepted
Funded PhD Project (European/UK Students Only)
The Image recorded at the sensor level looks nothing like the final photographs we see (and share). There are a number of processing steps – including white balancing [i], colour correction [ii] and tone mapping [iii] - which converts a raw image to the final output (for example jpeg). However, although people expect to see certain content in images to look a certain way – e.g. faces should be visible, snow should be white – this preference is not always delivered by current architectures which process all images in the same way (in essence process images in a content independent way).
This PhD project will consider how content can be determined in images (e.g. using deep learning [iv] architectures) and how processing pipelines can be adapted to make use of this content, including from an image fusion perspective [v]. There will also be an emphasis on developing approaches that can feasibly be implemented in embedded architectures (such as smart phones).
The project runs in collaboration with Spectral Edge Ltd, a Cambridge based company specialising in image processing. The PhD student will collaborate with researchers in Spectral Edge and spend some time in the Cambridge office (it is expected there will be internship opportunities).
For more information on the primary supervisor for this project, please go here: https://www.uea.ac.uk/computing/people/profile/g-finlayson
Type of programme: PhD
Start date of project: October 2018
Mode of study: Full time
This PhD studentship is jointly funded for three years by Faculty of Science and Spectral Edge. Funding comprises home/EU fees, an annual stipend of £14,296 and £1000 per annum to support research training. Overseas applicants may apply but are required to fund the difference between home/EU and overseas tuition fees (in 2017/18 the difference is £10,605 for the School of Computing Sciences but fees are subject to an annual increase).
Acceptable first degree: Computer Science, Physics, Mathematics, Engineering or other numerate discipline
Standard minimum entry requirement is 2:1
i) Finlayson, GD, “Correct-Moment Illuminant Estimation,” IEEE International Conference on Computer Vision, 2013
ii) Andersen, C and Connah, D, “Weighted Constrained Hue-Plane Preserving Camera Characterization. , 2016
iii) Qiu, G et al., “Learning to Display High Dynamic Range Images,” Pattern Recognition, 2007
iv) Peng. Et al, “Large Kernel Matters -- Improve Semantic Segmentation by Global Convolutional Network”, IEEE International Conference on Computer Vision and Pattern Recognition, 2
v) Connah, D et al. “Spectral Edge: gradient-preserving spectral mapping for image fusion,” Journal of the Optical Society of America A, 2016.