Abstract: A new algorithm for effective sub-pixel image features detection based on facet model is proposed. Based on the idea that the discrete image intensity function can be considered as sampled and noisy approximation of the underlying continuous or piecewise continuous image intensity function, a piece continuous function estimate called facets are used to represent the gray level intensity of the image pixel, which can be quickly gained by fitting a quadratic or cubic approximation surface on the constructed two-dimensional discrete chebyshev polynomials basis. The high precision image features of feature point and edge can be located with directional sec-ond-derivative zero-crossing operator on a local maximal gra-dient direction on the fitting facet surface. The image noise effect can be greatly reduced by using the image gray surface fitting method instead of the image interpolation, and can effi-ciently reduce the computational time. Experiments on many real images of a sub-pixel feature point detector and a sub-pixel edge detector are carried out, the results show that this method is stable, effective and precise localization, this sub-pixel fea-ture detector can effectively improve the precision of the vision calibration and measurement system.

Key words: Facet model, Computer vision, High precision image feature location