Face retrieval by an adaptive Mahalanobis distance using a confidence factor

Component-based LDA face description for image retrieval and MPEG-7 standardisation

“…The cascaded LDA reduces the matching complexity and the size of description for the combined LDA. Table 9 summarizes the comparitive results of the description techniques [2][3][4]8,22] proposed for the MPEG-7 meeting in May 2002. It is noted that the proposed component-based LDA method dramatically enhanced the retrieval performance of the previous methods.…”

“…False Rejection [1] The second-order eigenface method [2] 0.478 N/A The Fourier spectral PCA-based face description (with a confidence factor) [3] 0.243 N/A The eHMM with the second-order eigenvectors [4] 0.495 N/A The Pseudo2D-HMMs [8] N/A 0.554 The component-based LDA [22] 0.0678 0.1355 proposal submitted to the International Standardization body [20] in 2004 was successful.…”

“…To compensate for image variation due to illumination changes, Wang and Tan [2] proposed the second-order Eigenface method and Kamei and Yamada [3] extended their work to use a confidence factor describing face symmetry and intensity variation due to illumination change. Kim et al [5] developed the second-order PCA (Principal Component Analysis) Mixture Model (PMM) method.…”

“…Recently, face descriptors as candidates for MPEG-7 (Moving Picture Experts Group) standardisation have been proposed for face image retrieval in video streams [1][2][3][4][5][6]8,14,22]. The various methods developed for multimedia description and retrieval involve many kinds of visual factors like shape, color and motion.…”

A network of sensor-based framework for automated visual surveillance

“…The decision is taken using the Mahalanobis distance. Mahalanobis distance is a technique to determine similarity between a set of values and an unknown sample (Kamei, 2002). The Mahalanobis distance takes the information of the variance and covariance between variables.…”