Creation of image models for evolving objects on dynamically changing scenes
Yury Ipatov, Alexandr Krevetsky, Yury Andrianov, Boris Sokolov · Istrazivanja i projektovanja za privredu · 2017
An overview of the theoretical approaches to the formation of models images by key features. Dynamically changing scene images with selected objects of interest and background is reviewed. The statistical characteristics of this class of images are explored. Model convex hulls sampling distributions background / object in different colorimetric spaces were obtained. Synthesized model ellipsoids based on them using optimization methods give a good score for approximate calculation. Identified areas for further research and the use of the current results. Estimates of segmentation quality for algorithms based on synthesized models are calculated.