Using kernel principal components for color image segmentation

Slawo Wesolkowski

doi:10.1117/12.453858

21 November 2002 Using Kernel Principal Components for Color Image Segmentation

Slawo Wesolkowski

Proceedings Volume 4790, Applications of Digital Image Processing XXV; (2002) https://doi.org/10.1117/12.453858
Event: International Symposium on Optical Science and Technology, 2002, Seattle, WA, United States

Abstract

Distinguishing objects on the basis of color is fundamental to humans. In this paper, a clustering approach is used to segment color images. Clustering is usually done using a single point or vector as a cluster prototype. The data can be clustered in the input or feature space where the feature space is some nonlinear transformation of the input space. The idea of kernel principal component analysis (KPCA) was introduced to align data along principal components in the kernel or feature space. KPCA is a nonlinear transformation of the input data that finds the eigenvectors along which this data has maximum information content (or variation). The principal components resulting from KPCA are nonlinear in the input space and represent principal curves. This is a necessary step as colors in RGB are not linearly correlated especially considering illumination effects such as shading or highlights. The performance of the k-means (Euclidean distance-based) and Mixture of Principal Components (vector angle-based) algorithms are analyzed in the context of the input space and the feature space obtained using KPCA. Results are presented on a color image segmentation task. The results are discussed and further extensions are suggested.

Citation Download Citation

Slawo Wesolkowski "Using Kernel Principal Components for Color Image Segmentation", Proc. SPIE 4790, Applications of Digital Image Processing XXV, (21 November 2002); https://doi.org/10.1117/12.453858

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available