الفهرس 
Chapter1IntroductionOnly 14 pages are availabe for public view
 |  | from | 75 |  |  |
| | | from | 75 | | |
Abstract
Extracting foreground objects from still images or video sequences plays an important role in many image and video editing applications. Matting refers to the problem of accurate foreground estimation in images and video sequences; for image matting it’s easy to add some scribbles by the user manually, but for video matting it’s hard to add such scribbles manually by the user; background subtraction algorithms was used for this task after a lot of modifications.
Background subtraction in a video sequence is a commonly used class of techniques for segmenting out objects of interest in a scene for applications such as surveillance. If the scene is stationary or gradually evolving, foreground detection can be solved conveniently with many traditional background subtraction algorithms.
The present work investigates the performance of a number of background subtraction algorithms and video matting techniques. Moreover, improved versions of three existing algorithms are proposed. These are: frame difference, approximated median, and Mixture of Gaussian.
A new algorithm for background subtraction has been introduced and its performance was investigated. It is a combination of three background subtraction algorithms; frame difference, approximated median, and Mixture of Gaussian. The improved version of each algorithm is used and then a decision level fusion algorithm is applied to the three outputs.
An improved method of adding scribbles process is also proposed to minimize the percentage error of adding scribbles process and the results are satisfactory.