Clausius Normalized Field for Object Motion Analysis

This book introduces a Clausius Normalized Field (CNF) based on a novel homology between thermodynamic systems and images for the treatment of time-varying imagery, and we also present CNF modeling methods for motion segmentation and uncalibrated stereo matching problems. A CNF is a probabilistic model which reckons entropy variations by observing entropy definitions of Clausius and Boltzmann. A system colder than its surroundings absorbs heat from the surroundings, and the absorbed heat increases the entropy of the system, according to the entropy definition of Clausius. The increased entropy is also highly related to the disorder of the system as given by the entropy definition of Boltzmann. Because the pixels of an image are viewed as a state of lattice-like molecules in a thermodynamic system, reckoning the entropy variations of pixels is similar to estimating the degrees of disorder of the pixels.

Autorentext

Eunjin Koh received his Ph.D. degree from University of Science and Technology, Daejeon, South Korea in 2011. He is a research scientist at ADD (Agency for Defense Development), Daejeon, South Korea since 2011. His main research interests include image processing, human-computer interaction, object tracking, and pattern classification.