« IEEE fyrirlestur: CMS – Framtíðarlausn í þráðlausum mælingur frá Controlant | Main | IEEE fyrirlestur: Computational Intelligence and AI in Games: A New IEEE Transactions »
Contextual unmixing of geostatistical data based on Gaussian mixture distributions and Markov random fields
Skrifað af karl | February 10, 2009
Tuesday 17. February at 16:40 in room VR-2-155
Dr Nishii, Ryuei will give a talk entitled “Contextual unmixing of geostatistical data based on Gaussian mixture distributions and Markov random fields”
Abstract
Suppose that we are required to estimate fractions of categories in each mixel in a low-spatial resolution multispectral image. Nishii et al.
(IGARSS, 2008) proposed unsupervised contextual unmixing methods based on Gaussian mixtures for the feature vectors and MRF for the fraction vectors. We will extend the unsupervised contextual unmixing method into the case with a small training data set. This situation arises frequently because collection of training data is expensive and time-consuming. But, the training data are very helpful to obtain an appropriate classification even if they are small. Hyperdimensional case will be also discussed.
Topics: Óflokkað | No Comments »