Nonlinear dynamic texture analysis and synthesis using different types of KPCA

In nature, the texture is an omnipresent visual experience. The dynamic texture contains chromatic, spatial and temporal correlationships. Dynamic textures are categorized into linear dynamic texture and Nonlinear dynamic texture. Several methods have been employed to model dynamic textures like Sin...

Full description

Saved in:

Bibliographic Details
Published in:	2015 International Conference on Information Processing (ICIP) pp. 739 - 744
Main Authors:	Ghadekar, Premanand P., Parmar, Nilesh A., Chopade, Nilkanth B.
Format:	Conference Proceeding
Language:	English
Published:	IEEE 01-12-2015
Subjects:	Dynamics Gaussian kernel Heuristic algorithms HOSVD Kernel KPCA Mathematical model Nonlinear Nonlinear dynamical systems Principal component analysis simple kernel SVD
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	In nature, the texture is an omnipresent visual experience. The dynamic texture contains chromatic, spatial and temporal correlationships. Dynamic textures are categorized into linear dynamic texture and Nonlinear dynamic texture. Several methods have been employed to model dynamic textures like Singular value decomposition (SVD), Higher Order Singular value decomposition (HOSVD), and novel method of Fast Fourier Transform (FFT) with SVD. However, these methods are not able to capture nonlinear motion from dynamic textures with less model size. To model nonlinear dynamic textures Kernel Principal Component Analysis (KPCA) with its three types of kernel functions: Simple kernel, Polynomial kernel, and Gaussian kernel are used. Experimental results show that KPCA outperforms SVD approach. Among three kernels simple kernel and Gaussian kernel performs almost same but the time complexity of simple kernel is much less.
DOI:	10.1109/INFOP.2015.7489480