A comparison of resolution requirements for boundary estimation from binary and grey scale images

The determination of the appropriate sampling rate for image data is often problematic because the traditional band-limited reconstruction approach for data sampled above the Nyquist frequency cannot be applied to sources which are not band-limited. High contrast edges play an important role in image content definition, and these sharp edges in an image source would theoretically require an unbounded spatial frequency range. The paper presents a theoretical argument supported by several examples for selecting a different minimum sampling frequency for grey scale and binary image acquisition. Implications with respect to image acquisition, image processing, and image compression are discussed