A Taxonomy and Evaluation of Dense Two-Frame Stereo Correspondence Algorithms
International Journal of Computer Vision
Novel view synthesis in tensor space
CVPR '97 Proceedings of the 1997 Conference on Computer Vision and Pattern Recognition (CVPR '97)
Stereo Vision for View Synthesis
CVPR '96 Proceedings of the 1996 Conference on Computer Vision and Pattern Recognition (CVPR '96)
High-quality video view interpolation using a layered representation
ACM SIGGRAPH 2004 Papers
PET'02 Proceedings of the 2nd international conference on Privacy enhancing technologies
Novel view synthesis for stereoscopic cinema: detecting and removing artifacts
Proceedings of the 1st international workshop on 3D video processing
Exposing digital forgeries in color filter array interpolated images
IEEE Transactions on Signal Processing - Part II
| Hi-index | 0.00 |
In recent years, social media has played an increasingly important role in reporting world events. The publication of crowd-sourced photographs is one of the reasons behind the high impact. However, the use of a camera can draw the photographer into a situation of conflict. Examples include citizen journalists posting photographs of incidents of human rights violations on the Internet. The published images contain unambiguous clues about the location of the photographer such as the angle from which the scene was captured. Further, in the context of adversary operated CCTV systems, knowledge of the photographer's potential location allows the adversary to identify the photographer by reviewing relevant footage. This is the camera location detection attack --- a novel privacy threat against photographers seeking anonymity while posting images. In order to resist such powerful attacks, we introduce the notion of camera location anonymisation; by combining multiple input images captured from different viewpoints, we produce a single image that appears to have been shot from a randomly chosen angle. To this end, we examine the use of view synthesis algorithms from computer vision literature as concrete defences. We show (both analytically and experimentally) that such defences could be a promising step in the direction of providing probabilistic anonymity guarantees. We analyse the extent of anonymity such techniques can provide in various scenarios and describe the challenges posed by scene geometry.