Non-parametric Local Transforms for Computing Visual Correspondence
ECCV '94 Proceedings of the Third European Conference-Volume II on Computer Vision - Volume II
Real-Time Image-Based Motion Detection Using Color and Structure
ICIAR '09 Proceedings of the 6th International Conference on Image Analysis and Recognition
An evaluation framework for stereo-based driver assistance
Proceedings of the 15th international conference on Theoretical Foundations of Computer Vision: outdoor and large-scale real-world scene analysis
| Hi-index | 0.00 |
This paper presents a new approach to motion segmentation. Unlike most other methods, the proposed approach relies on summarized local image structure obtained by census transform, and not on absolute luminance values. We apply census transform for two frames from the sequence which are registered to the sub-pixel accuracy regarding to the moving target object. The registration is carried out by estimation of affine parameters. The two sets of resulting local image structures, in the forms of bit strings, give good and dense correspondence on the object to segment the region. We study the effectiveness of the sub-pixel registration compared to the pixel unit registration. We show segmentation results for natural scenes.