Undergraduate Thesis: Hyperacuity Sensing for Image Processing

The human eye’s response to high spatial frequency content in images is highly dependent on the nature of the features: the eye integrates periodic intensity variations at high frequencies to form a solid gray, yet is very sensitive to high-contrast steps in intensity (which contain high frequency components), such as edges. This study presents the development and effectiveness of a scanning system that has a similarly varying response to spatial frequency content. The system is implemented by an array of PSD (position sensitive detector) scanning elements, each of which produces four values which can be composed to retrieve intensity and intensity gradient information over the detector’s area. It is shown that this information, along with additional information from immediately neighboring pixels, can be used to determine the frequency content and edge locations of the image area covered by the detector. The accuracy of this determination is shown to vary in a way similar to that of the human eye’s response. A singlepass algorithm (BBJJ) is developed which determines how to render the area of the image corresponding to each detector as either an edge or a smooth gradient. A version of the algorithm is also presented for standard scanning detectors, which composes the information from multiple detectors to construct values analogous to those available from each PSD element. The algorithms described were implemented in an existing simulation framework. The performance of each of these methods in comparison to that of a standard scanning system is shown (in simulation) for three major classes of images: line art, continuous tone, and halftoned images.