Adaptive Object Selection Based on Multi-Threshold Image Processing

Adaptive object selection technique based on the results of multi-threshold image processing is considered. The keynote feature of the proposed approach is the use of information about the properties of the selected object(s) at multiple threshold values a posteriori to find an optimal threshold based on the criterion that takes into account the geometric properties of the object. Multi-threshold object selection algorithms based on several scale-invariant geometric criteria such as the object perimeter elongation coefficient and the elongation coefficient of the circumscribed ellipsoid main axis are implemented and validated. The suggested characteristics help to distinguish between compact and elongated objects while remaining invariant to the image scale. The results for both synthetic and observational imagery obtained by remote sensing equipment are used for validation purposes.