K.U.Leuven > ESAT > PSI > Visics > Research > Topics > Item 3.3

Image processing operations for forensic support (IMPROOFS)

L. Van Gool, J. Schouteden, T. Tuytelaars, M. Pollefeys

This European project aimed at the development of tools for forensic image processing. Increasingly, (surveillance) cameras are the silent witnesses of crimes. These data may contain more information than what is obvious to the naked eye. IMPROOFS has in particular focused on three kinds of image analysis: 1) image enhancement, 2) metrology, and 3) person identification through gait analysis or 3D mug shots. For the first topic, non-linear partial differential equations have been developed, that yield enhanced images where noise is reduced and simultaneously edges and structures are sharpened. Also super-resolution techniques were developed, that result in higher resolution images, taking overlapping pictures from a video sequence and combining them into 1 picture. Another set of tools allowed to subtract periodic and non-periodic backgrounds from images that contain interesting patterns such as fingerprints. The second strand included research into the estimation of the height of people from a single view, or the reconstruction of a scene's layout from two or more images of a scene with known parallel or orthogonal structures. The third topic included work on the analysis of gait patterns from video data, with the goal of differentiating between the patterns of different people. Another part of the work was the development of a simple and low-cost 3D face modelling set-up that allows taking 3D photographs of suspects. These allow to better compare the faces of suspects against those of offenders visible in surveillance images. Indeed, the face can be rotated and scaled at will, and different illumination conditions can be simulated. The partners were the Forensic Science Services (UK), INRIA (F), KTH (S), the National Institute for Criminology and Criminalistics (B), and the Universities of Leuven (B, coordinator) and Oxford (UK). More information about the project can be found on the project home page.

Single original image	Measurement of person height