Recent advances in deep learning for tomographic reconstructions have shown great potential to create accurate high quality images with a considerable speed-up compared to classical reconstruction methods. This is especially true for model-based learned (iterative) reconstruction schemes. However, applicability to large scale inverse problems is limited by available memory for training and extensive training times.
In this talk I will discuss applicability of learned image reconstruction approaches to tomographic data. In particular we will discuss various imaging scenarios and modalities, suitable approaches to design a robust learning task, as well as some solutions to obtain scalable learned image reconstruction for large scale and high dimensional data.
- Speaker: Professor Andreas Hauptmann
- Friday 30 August 2019, 14:00–15:00
- Venue: MR11, Centre for Mathematical Sciences, Wilberforce Road, Cambridge.
- Series: CMIH seminar series; organiser: J.W.Stevens.