Sensor interpolation and super resolution image reconstruction using the optimal recovery framework
Wireless sensor networks have the potential to benefit society in a myriad of ways, but due to constraints of sensor placement, obtaining the desired signals is not an easy problem. Retrieving the information accurately from sensors placed randomly is a challenging problem of sensor communication and signal interpolation. In this research the optimal recovery (OR) method, a deterministic framework that can use a priori bandwidth or spectral shape information, is used to interpolate from the given samples produced by sensors. In the reconstruction formulation, a worst case error signal occurs. To minimize this error, the OR method uses a norm to measure deviation over all possible feasible signals. In this thesis, monitoring of worst case errors serves to assess sensor deployment configuration quality and to optimize the placement of additional sensors. Optimal deployment of sensors and random deployment of sensors are two ways of positioning additional sensors. These two approaches are compared and contrasted to show the improvement that is possible using the OR framework. This method can also be used in the reconstruction of two-dimensional images in the super resolution (SR) problem. When multiple images are taken from the same scene in a short span of time, this results in spatially shifted low resolution images which are the inputs in the SR problem. The OR method is used here in the reconstruction of the output high resolution image on a block-by-block basis. The effects of errors and reconstruction problem sensitivity images are reduced using a regularization parameter. A recently developed Cross validation approach is used to compute an optimal regularization parameter. Comparison with other SR methods is done to show the potential superiority of the OR based SR reconstruction.
Moram, Veenarai, "Sensor interpolation and super resolution image reconstruction using the optimal recovery framework" (2010). ETD Collection for University of Texas, El Paso. AAI1484160.