Renaissance of the Sampling Theory

About project

Signal sampling and reconstruction are a central part of the digital signal processing (DSP) - the sampling theorem provides a bridge between the continuous and discrete-time world in modern electronic devices, smartphones, cameras and medical equipment. Compressive sensing (CS) is a technique that relies on finite rate of innovation in order to reduce the number of measurements needed for statistically reliable reconstruction. Rapid advances in this research area bring a renaissance of the sampling theory, mostly due to its significance to the DSP community.

Our research contributes to this challenging field in several aspects. New post-sampling paradigm will be based on hybrid continuous-discrete data models and processing methods. It will enable continuous signal processing on a digital computer without the discrete approximation based on samples.

Moreover, novel optimal and adaptive measurement approaches will be introduced. Exploratory measurements will be proposed to overcome certain drawbacks of the CS dimensionality reduction. Recently, a deep learning based approach resulted in a rapid advance of supervised learning methods. We will apply machine learning methods to achieve high quality and fast CS reconstruction.

Applications that will benefit from our research results are multi-view and light field imaging, 3D reconstruction and medical imaging, namely positron emission tomography (PET). Light field imaging goes beyond the conventional photography and enables powerful post-capture capabilities, such as change of viewpoint, digital refocusing, and virtual change of aperture. Our research will enable more accurate 3D reconstruction from multi-view, light field and PET images.

Team members

prof. dr. sc. Damir Seršić
project leader
full professor,  Department of Electronic Systems and Information Processing