1
|
Abstract:
We describe and analyze a new algorithm for performing efficient classification on large two-dimensional data sets, typically large digital images. RewriteWe use non-parametric classifiers on a lower resolution representation of the data set, namely the lowest subband of its Discrete Wavelet Transform decomposition. In this representation, each sample corresponds to a block of samples from the original data. At each step of the classification process we decide to either classify the whole block as belonging to a certain class, or to re-examine the data at a higher-resolution level, by moving down one level in the wavelet decomposition pyramid. In the parametric case, our analysis shows that, compared to traditional sample-by-sample classification techniques, [3, 4, 5] this new progressive scheme is not only more efficient (in terms of the number of operations required), but also, under plausible conditions, it is more accurate in its results. Experimental results illustrating this performance on the classification of large satellite images are reported in an earlier paper [2].
Citations
| 2961 | Pattern Classification and Scene Analysis – Duda, Hart - 1973 |
| 182 | Wavelets and filter banks: Theory and design – Vetterli, Herley - 1992 |
| 181 | Wavelets and signal processing – Riol, Vetterli - 1991 |
| 31 | A discrete-time multiresolution theory – Rioul - 1993 |
| 23 | Random Fields on a Network – Guyon - 1995 |
| 20 | A fast k nearest neighbor finding algorithm based on the ordered partition – Kim, Park - 1986 |
| 9 | Design of neural networks for classification of remotely sensed imagery – Chettri, Cromp, et al. - 1992 |
| 8 | Data mining of multidimensional remotely sensed images – Cromp, Campbell - 1993 |
| 6 | Probabilistic neural network architecture for high-speed classification of remotely sensed imagery – Chettri, Cromp - 1993 |
| 1 | and Ioannis Kontoyiannis, "Progressive Classlification in the Compressed Domain for Large EOS Satellite Databases – Castelli, Li, et al. - 1996 |
