Identification of Institutional Logo based on Wavelet Features

Abstract

The main aim of this paper is to recognize the logos of the input document so as to process the document for its classification and analysis, an algorithm is proposed using texture features based on Discrete Wavelet Transform (DWT) and Fast Fourier Transform (FFT) of object occurrence in a new tessellation of logo images and these features are given to the SVM and KNN classifier for recognizing the logos. The proposed algorithm is experimented on a data set of Institutional logos. The experimental results have shown the average recognition accuracy as 67. 74% using NN classifier, 79. 35% using KNN classifier and 87. 09% using SVM classifier. It is an initial attempt towards the classification of documents based on logos.

Cite this paper

@inproceedings{Dhandra2014IdentificationOI, title={Identification of Institutional Logo based on Wavelet Features}, author={Basanna V. Dhandra and Shridevi Soma and Gururaj Mukarambi and David S. Doermann and Ehud Rivlin and Ittay Weiss and Rashmi T and Gururaj and Mallikarjun Hangarge and Ju-Bing Chen and Mun K. Leung and Sina Hassanzadeh and Hossein Pourghassem and Kamarul Hawari Ghazali and Mohd. Marzuki Mustafa}, year={2014} }