The human visual system processes the retinal light distribution with different resolutions. Thresholds exceeding the accuracy of the sampling mosaic by an order of magnitude are obtained in a class of spatial hyperacuity tasks. Considering an optimal estimation model, we show that a priori information derived from a coarse map of the input does improve the spatial hyperacuity performance in certain tasks. An algorithm is presented for subpixel precision recovery from discrete images of elongated features' orientation.
|Original language||English (US)|
|State||Published - 1985|
All Science Journal Classification (ASJC) codes