Title: Content based image retrieval with multi-channel LBP and colour features

Authors: M.O. Divya; E.R. Vimina

Addresses: SCMS Cochin School of Business, Prathap Nagar, Mutttom P O, Aluva – 683106, India; Department of Computer Science and IT, Amrita School of Arts and Sciences, Amrita Vishwa Vidyapeetham, Kochi Campus, India ' Department of Computer Science and IT, Amrita School of Arts and Sciences, Amrita Vishwa Vidyapeetham, Kochi Campus, India

Abstract: Content based image retrieval (CBIR) systems are used for retrieving relevant images from datasets in response to a query based on image features. The features used for describing the image content play crucial role in determining the efficacy of the CBIR. In this paper a texture-colour fusion method exploiting the multi-channel information of colour images is proposed for describing the image content. Texture is represented with multi-channel local binary adder pattern, computed by considering all the channels of a colour image, and colour information is computed by quantising the constituent colour channels of the image. The method exploits RGB colour space for feature extraction. Experimental results show respective average retrieval precisions of 80.73%, 60.096%, 48.22% and 69.89% in the Wang's, Corel 5k, Corel 10k and Zubud datasets using the proposed feature combination. Comparative analysis indicates that the proposed approach has an edge over many other recent methodologies under consideration.

Keywords: multi channel LBP; colour quantisation; feature normalisation; image similarity; image retrieval; local binary patterns; LBP; feature fusion; RGB; colour channels; precision; recall.

DOI: 10.1504/IJAPR.2020.111524

International Journal of Applied Pattern Recognition, 2020 Vol.6 No.2, pp.177 - 193

Received: 13 Dec 2019
Accepted: 21 Jun 2020
Published online: 30 Nov 2020 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article