EFFECT OF POINT-SPREAD FUNCTIONS ON THE EDGE ORIENTATION PRECISION OF 1ST-DERIVATIVE OPERATORS

We examine how different imaging sensor point-spread functions (PSFs) affect the precision to which small-template first-derivative operators can measure step-edge orientation. Three PSF models are tested under different numbers of image quantization levels. For each model, the precisions produced by several first-derivative operators are computed using different error measures. Optimal estimates derived using the theory of locales are also provided as a basis for comparison. It is shown that the size of the PSF significantly affects the precision to which orientation can be found. Precision is highest when the PSF has a spread of 1.0 to 1.4 pixels and is dependent on the PSF shape. A method to produce optimal first-derivative templates for specific PSFs is also described.