Back to Projects | Mushfiqur's Homepage
Ghost removal is a challenging problem in merging multiple exposures into an HDR image. While camera movements can be compensated by alignment and registration algorithms, object movements within the scene can cause significant ghost artifacts if not handled properly.
Debevec and Malik [1] presented the first algorithm to compute the HDR image and the response function of a camera from a sequence of exposures with different exposure times. Robertson et al [2] gave a more robust method in 1999. However, most of this algorithms assume the scene is perfectly static. The algorithms break down if the scene is not static and produces ghost artifacts.
It has been observed that a reference image has to be used in order to remove ghost artifacts. Recent works act on three images [3]; but in most cases, three images are not enough to capture the whole dynamic range. In this project, I tried to address this problem.
In the images below, 4 different exposures are shown. The top row is the result of the algorithm developed in the project, the bottom row is generated by HDRgen [4]. The original image sequence has 12 images.
| This algorithm |
|
| HDRgen | |
| This algorithm |
|
| HDRgen | |
| This algorithm |
|
| HDRgen |
[1] P. E. Debevec and J. Malik, Recovering High Dynamic Range Radiance Maps from Photographs, Proceedings of International Conference on Computer Graphics and Interactive Techniques, pages 369-378, August 1997.
[2] Robertson, M.; Borman, S. and Stevenson, R., Dynamic range improvements through multiple exposures, Proceedings of International Conference on Image Processing, pages 159-163, 1999.
[3] Nicolas Menzel and Michael Guthe, Freehand HDR Photography with Motion Compensation, Vision, Modeling and Visualization, 2007.
[4] HDRgen, Command Line HDR Image generation software for Linux, http://www.anyhere.com