A subpixel algorithm using interpolation method based on normal direction

As one branch of subpixel technology, curve interpolation method is the most popular for its higher precision and less complexity. But the method has longer running time and bigger direction error, since it makes use of convolution between direction templates and pixel gray values to get Gradient direction which is interpolating direction. So a new method is proposed in which unisometric cubic spline interpolation is operated along the normal direction of the edge-point. It illustrates not only how to choose sampling points for interpolation, but also how to figure node values of those points, then establishes the suitable interpolating boundary conditions. The method reduces not only the uncertainty but also the runtime to 1/2, and improves measuring precision and efficiency.