Quantitative Analysis of Resolution Synthesis (TA-P7)

Author(s) : Ramez Yoakeim (University of New South Wales, Australia) David Taubman (University of New South Wales, Australia)

Abstract : We address a number of gaps left by recent work on resolution synthesis image interpolation, including the validity of some assumptions which we examine and verify. The relative merit of a mixture approach versus a maximum likelihood approach, the impact of the classification process, and the impact of the number of classes on the performance of the interpolator are all investigated. We also examine the suitability of the underlying statistical models. We propose a modified synthesis component, based on the discrete wavelet transform, as an alternative to the non-overlapped block synthesis process described in previous work. We also introduce a number of measures to significantly improve the computational efficiency and suitability of RS for automated, unassisted classification training.

Menu