Department of Information Technology
Department of Information Technology
Uppsala UniversityInformation TechnologyResearch
Listen

Image analysis

An image says more than a 1000 words

Overview

Computerised image analysis is about developing computational methods for extracting meaningful information from images - mainly from digital images - by means of digital image processing techniques, including for example convolutional neural networks (CNNs).

We develop theory, methods, algorithms, and systems to address questions related to life science, medicine, digital humanities, and other applications. This include identifying objects, extracting measurements, and making decisions based on image data. Many methods are common for wide ranges of applications. The list of research topics below is therefore related to several of our research entities and projects.

Research Topics

Over the years, more and more of the research in image analysis involves development and application of model- and learning-based (AI) methods. However, some traditional methods also necessarily remain and are developed.

  • Image reconstruction and de-noising includes methods for improving the quality of image data.
  • Image registration addresses methods to computationally align image data collected e.g. at different time points or with different imaging modalities.
  • Digital geometry focuses on deriving geometric information from digital images, taking the limitations of discrete representations into account.
  • Object detection can broadly include both delineation and classification of objects and images.
  • Feature extraction includes approaches to extracting measurements and other properties from objects or regions of interest in images, relevant for the subsequent analysis.
  • Image understanding is the ultimate goal of image processing and analysis, and provides interpretation of the information contained in the image data.
  • Visualization is any technique for creating images, diagrams, or animations to communicate a message - make the invisible of scientific data visible.
  • End-to-end image analysis proposes deep learning-based methods optimized to directly interpret image data fed into the system, without performing any intermediate steps.
Updated  2023-11-20 20:12:26 by Ekta Vats.