Probabilistic Image Modelling

Lead Research Organisation: University College London
Department Name: Computer Science

Abstract

To study applications of, and improvements to, probabilistic modelling of images. In particular the application towards lossless compression, which is poorly understood by the research community, and fundamental improvements in image modelling techniques. These applications and techniques are built around using neural networks to model the images.

Publications

10 25 50

Studentship Projects

Project Reference Relationship Related To Start End Student Name
EP/R512400/1 01/10/2017 31/03/2022
1926563 Studentship EP/R512400/1 25/09/2017 04/08/2022 Thomas Bird
 
Description We have improved the state-of-the-art for lossless image compression through machine learning. We have in principal shown that a system trained on data can compress images to smaller sizes than any existing systems, in a practical manner.
Exploitation Route Image compression is highly relevant for many other fields, since data is abundant these days and we are looking to reduce the footprint it produces.
Sectors Digital/Communication/Information Technologies (including Software)

 
Description These findings are beginning to be used and investigated by industry, namely large technology companies who have an interest in furthering lossless image compression.
Sector Digital/Communication/Information Technologies (including Software)
Impact Types Economic

 
Title Craystack: an open source python package for building lossless codecs 
Description A python package for prototyping lossless compression algorithms 
Type Of Technology Software 
Year Produced 2019 
Open Source License? Yes  
Impact This was the byproduct and enabled publications in the area, and will further help lossless compression research. 
URL https://github.com/j-towns/craystack