AG 1, AG 2, AG 3, AG 4, AG 5, SWS, RG1, MMCI   Info

Public Audience

English

120 Minutes

Saarbrücken

High Dynamic Range (HDR) imaging aims at replicating real world scenes faithfully by compositing multi-exposure images using the knowledge of camera response function (CRF). The generated HDR image needs to be tone-mapped for compatibility with common displays and printers. Alternately, one can directly generate a low dynamic range (LDR) image from multi-exposure images even if CRF is not known. This approach poses a problem in the case of dynamic scenes where one needs to detect and eliminate moving objects in the scene before compositing. We propose a unified gradient domain approach to generate artifact-free LDR images for dynamic scenes by compositing their multi-exposure images without the knowledge of CRF. The primary contribution of our work is to develop comparagram based method for blind detection of motion patches in the absence of any knowledge of camera parameter settings. Additionally, the multi-exposure compositing is performed in the gradient domain which prevents seams in the output LDR image.

A common method for real-time segmentation of moving regions in image sequences involves “background subtraction”. Background subtraction as proposed by Stauffer-Grimson models pixel intensity as a mixture of Gaussians. The Gaussians’ mean and variance adapts continuously in accordance with observed intensity at a given pixel of the image sequence. Since this method has a limited tolerance towards sudden intensity variations at a particular foreground/background pixel, it fails to model intensity variations induced by effects such as Camara’s Automatic Gain Control (AGC). We propose an enhancement to the Stauffer-Grimson method where we model such intensity variations robustly. Camera AGC is estimated with the aid of comparagram between consecutive frames of a video sequence

IMPRS Office Team

0681 93251800

--email hidden

passcode not visible

logged in users only