Abstract
In digitizing images or signals, the results after A/D conversion are always quantized. Quantization will introduce uncertainty in image/signal values as compared to the original continuous-distributed image intensity or value of the physical signals, and will sometimes cause loss of important information, deterioration of image/signal quality and thereafter affect the effective useages of the image/signal data. In this thesis, we present a class of point-processing techniques to reduce the uncertainty of image/signal value caused by quantization. They utilizes several digitized images acquired via varying the transmittances of the inserted media, aperture or A/D conversion, and offer ways to combine these different images to provide a new image of finer quantization accuracy. Experimental results from tests on images show that these techniques do reduce quantization error greatly and provide an image of much better quality.