Antinoise texture retrieval based on PCNN and one-class SVM

Abstract

By training and predicting the features that are extracted by pulse coupled neural network (PCNN), a noise immunity texture retrieval system combined with PCNN and one-class support vector machine (OCSVM) is proposed in this paper, which effectively improve the anti-noise performance of image retrieval system. The experiment results in different noise environment show that our proposed algorithm is able to obtain higher retrieval accuracy and better robustness to noise than traditional Euclidean distance based system. © 2013 Springer-Verlag Berlin Heidelberg.