direkt zum Inhalt springen

direkt zum Hauptnavigationsmenü

Sie sind hier

TU Berlin

Page Content

Real-time Image Dehazing - Christian Thurow

 Our atmosphere contains huge quantities of particles, compromising outdoor photogra- phy. They cause scenes to appear hazy or foggy, this reduces visibility of objects and their contrast, and makes detection of objects within the scene more difficult. However, due to recent developments in computer vision, it is now possible to improve outdoor images and remove the haze layer. Many computer vision applications can benefit from haze free images. These techniques are physically sound and based on theories from me- teorology and other disciplines. Unfortunately, these techniques are so costly in terms of complexity that they are not suitable for real-time applications. In this thesis a method is proposed that is able to dehaze images in real-time, employing an algorithm with com- plexity of only O(n), whereas most existing algorithms inherit a complexity of O(n2). This algorithm is based on a strong, statistically based prior, the dark channel prior. For its evaluation, a new measure for the degree of dehazing is introduced. This thesis shows that real-time dehazing is possible. It’s validity and effectiveness is evaluated with video material from airport ground surveillance. These results enable real-time image dehazing for new applications like high resolution, high frame rate outdoor surveillance, on board vehicle camera applications and many more.

Download Thesis


Zusatzinformationen / Extras

Quick Access:

Schnellnavigation zur Seite über Nummerneingabe