With shadow elimination towards effective foreground extraction

Abstract

Very important process in the area - image processing, is the task of foreground extraction. To learn a computer to distinguish between foreground objects and background scene is non-trivial. Further, if foreground is extracted, a question can be set: is the foreground credible? What if significant part of detected foreground is unimportant information? Shadow is in most cases as noise considered. Our goal is to analyze possibilities shadow elimination and next to provide improved inputs of detected foreground for more efficient traffic surveillance system.