direkt zum Inhalt springen

direkt zum Hauptnavigationsmenü

Logo der TU Berlin

IV Topics in Computer Vision

Topics in Computer Vision

Semester: Sommersemester
Art: IV 4
LV-Nr.: 0435 L 726
Veranstalter: Swaminathan, Albayrak
VL: Donnerstag 10:00 - 12:00 im FR 3002
UE: Montag 10:00 - 12:00 im FR 3002
Beginn am 20.04.2006

Please visit the homepage of the course: http://www.cv.tu-berlin.de/lehre/cv/htcv/sem_swaminathan.phtml/

This course is held in English

Flyer zur Lehrveranstaltung

The last few years have seen a variety of new and exciting avenues for applying computer vision research. In this course I begin with traditional problems in computer vision to motivate these new areas and then explore recent advances in a handful of these research directions including (but not restricted to):

  • Novel Imaging Geometries: We start with the first class of generalization beyond the pinhole model to single viewpoint or central imaging systems. We cover the design criteria, as well describe the unifying models for single viewpoint catadioptric imaging systems. We then explore newer camera designs which break the bounds of single viewpoint imaging. We analyze these new geometries as well as present theory on the design of new general imaging systems. A project to design a new camera or projection system might be included.
  • Vision and the Weather: The image acquired is clearly affected by the current illumination and weather conditions. Effects such as fog, mist rain severely reduce visibility and image quality. However, most practical vision systems must cope with such changes in weather conditions. We shall study new approaches to model these weather effects using simple but highly effective techniques.
  • High Dynamic Range Imaging: The human eye is capable of seeing a large range of intensity values within any seen. This is typically considered to be millions of intensity values in terms of pixel intensities. The traditional camera however does not possess such a large "dynamic range". Typically the camera only has 8 bits per colour channel providing at most 256 intensity values. In this section we shall discuss many methods and current research cameras that try to extend the dynamic range of cameras. Such techniques, allow us to capture images which contain bright outdoor scenes as well as indoor shadowed darker regions within a single image.

Part one includes theory of single and non-single viewpoint imaging geometry, calibration, design of general imaging and projection systems. Part two will treat the problem of appearance changes due to weather effects. Single and multiple scaterring models and methods to remove fog, mist, haze will be discussed.

Voraussetzungen

  • Vordiplom

Prüfungsmodalitäten, Anforderungen

Diese Lehrveranstaltung kann in eine Prüfung in den Bereichen KI, BKS und WVA eingebracht werden.

Ansprechpartner