Background extraction, which is to compute a background image given a set of images, has real life applications in computer vision such as video surveillance and High Dynamic Range imaging. Moving objects in the provided set of images make the problem challenging. Obtaining such a background image is usually the first step in the aforementioned tasks and the performance is heavily influenced by the extracted background image.

Background extraction has been studied and different methods have been proposed to solve the tasks under various constraints. A difficult scenario is to extract the background with very few images. We proposed a simple framework that utilizes sampling and clustering techniques to solve the background extraction problem in this setting. Experiments are condu...[ Read more ]

Background extraction, which is to compute a background image given a set of images, has real life applications in computer vision such as video surveillance and High Dynamic Range imaging. Moving objects in the provided set of images make the problem challenging. Obtaining such a background image is usually the first step in the aforementioned tasks and the performance is heavily influenced by the extracted background image.

Background extraction has been studied and different methods have been proposed to solve the tasks under various constraints. A difficult scenario is to extract the background with very few images. We proposed a simple framework that utilizes sampling and clustering techniques to solve the background extraction problem in this setting. Experiments are conducted to evaluate the performance of the proposed method with real world datasets. We also qualitatively compared the results with several existing state-of-the-art methods.