The Community for Technology Leaders
Pacific-Asia Workshop on Computational Intelligence and Industrial Application, IEEE (2008)
Dec. 19, 2008 to Dec. 20, 2008
ISBN: 978-0-7695-3490-9
pp: 1007-1011
ABSTRACT
This paper presents a new no-reference perceptual blur metric by using the radial basis function network which is based on the orthogonal least squares learning algorithm (OLS-RBF). It extracts the generalized local features of the edge points in structure-texture region and acquires the generalized image features by performing Principal Component Analysis (PCA) on the average of generalized local features. The Gaussian blurred image quality estimation involves making the function relationship between the generalized image features and subjective scores. This paper transforms the problem of quality estimation to a problem of function approximation and solves the problem by using OLS-RBF network. OLS-RBF network uses an orthogonal least squares learning algorithm to select suitable centers for the radial basis function, which makes the training procedure simpler. Experiments results on various Gaussian blurred images show that the new metric's performance is consistent with the subjective evaluation and outperforms other blur metrics.
INDEX TERMS
CITATION
Zhang Hua, Zhu Wei, Chen Yaowu, "A No-Reference Perceptual Blur Metric by using OLS-RBF Network", Pacific-Asia Workshop on Computational Intelligence and Industrial Application, IEEE, vol. 01, no. , pp. 1007-1011, 2008, doi:10.1109/PACIIA.2008.173
83 ms
(Ver 3.3 (11022016))