Title :
Shadow identification
Author :
Jiang, Caixia ; Ward, Matthew O.
Author_Institution :
Worcester Polytech. Inst., MA, USA
Abstract :
A shadow identification and classification method for real images is developed. The method is based on extensive analysis of shadow intensity and shadow geometry. The procedure for identifying shadows is divided into low-level, middle-level, and high-level processes. The low-level extracts dark regions from images. The middle-level process performs feature analysis on dark regions, including detecting vertices on the outlines of dark regions, identifying penumbrae in dark regions, assigning the subregions in dark regions as self-shadows and cast shadows, and finding object regions adjacent to dark regions. The high-level process integrates the information derived from the previous processes and confirms shadows among the dark regions
Keywords :
image segmentation; pattern recognition; cast shadows; dark regions; detecting vertices; feature analysis; high-level process; real images; self-shadows; shadow classification; shadow geometry; shadow identification; shadow intensity; Clouds; Geometry; Histograms; Humans; Layout; Light sources; Object detection; Performance analysis; Shape; Surface texture;
Conference_Titel :
Computer Vision and Pattern Recognition, 1992. Proceedings CVPR '92., 1992 IEEE Computer Society Conference on
Conference_Location :
Champaign, IL
Print_ISBN :
0-8186-2855-3
DOI :
10.1109/CVPR.1992.223128