Urban scenes and structures often appear to be distorted in photos taken by casual photographers, as the scene layout often conflicts with how it is expected by human perception
Abstract
Urban scenes and structures often appear to be distorted in photos taken by casual photographers, as the scene layout often conflicts with how it is expected by human perception. In this project, we propose an automatic approach for straightening up slanted man-made structures, based on a set of criteria that were defined according to human perception studies. We’ve developed an optimization framework which yields an optimal homography for adjustment.
Robust automatic algorithm
- Various criteria for perceived quality
- Energy minimization framework
- Natural scene handling
- Fast and effective method
Flowchart
Adjustment criteria
Adjusting object’s lines to the squared frame
Eye level alignment
Perspective distortion
Image distortion
Results –urban scenes
original image our image
Results – indoor scenes
original image our image
Results – natural scenes
original image our image
Our result vs. regular rectification
Conclusions
- An automatic upright adjustment algorithm was implemented successfully
- The algorithm produces better results then regular rectification
- The proposed solution can be improved with dynamical photo classification and energy weights determination
- Execution time should be decreased for better user experience