Negative Coefficient Polynomial Kernel Density Estimation for Visualization

S.D. Witherspoon and M. Zhang (USA)


Visualization Simulation, NCoP Estimation, Object Extration


The Negative Coefficient Polynomial (NCoP) employs nonparametric Kernel Density Estimation (KDE) technique to post process images and to produce difference images and statistics that offer meaningful measures to determine the kernel function effectiveness in object extraction. In this paper, three NEW kernel functions of NCoP was developed (Hyperbolic Cosecant, Skewed Polynomial and Negative Polynomial kernel functions) to compare with commonly used kernel functions. Four experiments were designed to evaluate the KDE functions: 1) moving object 2) lighting change 3) moving background and 4) missing object. Results indicate that the polynomial functions yielded a 70% false detection rate compared to <5% for the other functions in the lighting change experiment. The Negative Polynomial outperformed all functions in the other experiments by averaging <3% false detection while achieving object extraction. Concluding, the Negative Polynomial produced better results for object extraction using KDE when the lighting was constant.

Important Links:

Go Back