Title :
Kernels in structured multi-feature spaces for image retrieval
Author :
Djordjevic, D. ; Izquierdo, E.
Author_Institution :
Dept. of Electron. Eng., Queen Mary Coll., London, UK
fDate :
7/20/2006 12:00:00 AM
Abstract :
A new kernel for structured multi-feature spaces is introduced. It exploits the diversity of information encapsulated in different features. The mathematical validity of the introduced kernel is proven in the context of a conventional convex optimisation problem for support vector machines. Computer simulations show high performance for classification of images.
Keywords :
convex programming; image classification; image retrieval; support vector machines; convex optimisation problem; image classification; image retrieval; structured multifeature spaces; support vector machines;
Journal_Title :
Electronics Letters
DOI :
10.1049/el:20061454
