Adaptive Extraction of Small Objects in Digital Images

Abstract

The problem of detection and localization of various size and shape small-extended objects in electronic surveillance systems using synthetic aperture radar, lidar, infrared and television cameras is discussed. An intensive and non-stationary background is described as the main difficulty in processing. This problem is solved using oriented filtering, adaptive thresholding and morphological analysis. Improved method is proposed for the adaptation of detection threshold based on the analysis of isolated fragments remaining in the image after thresholding.

Keywords

• фильтрация
• локализация объектов на изображениях
• адаптивная пороговая обработка