Analysis of the Quality of Thresholding for Object Detection in Images

Abstract

The quality of threshold processing for three algorithms for detecting and selecting objects of interest on location and television images obtained by radio surveillance systems and technical vision systems is studied. Algorithms form adaptive global and local thresholds using object selection and without it. The quality of threshold processing is evaluated on ROC curves and Recall-Precision characteristics by modeling using a Gaussian model, and is checked on a real image.