Sub-pixel Location Algorithm of Complex Edge Target Based on Gradient Direction

In the existed sub-pixel edge location methods, the quadratic curve fitting edge location is a better approach from the consideration of the complexity and computation time. However, conventional methods can only accurately locate the straight edges with specific slope, for which this paper proposes an improved sub-pixel edge location algorithm that can be used to detect the curve edge with the complex characteristic of changing tangent slops, and naturally can also detect the linear edge. The algorithm firstly locates the edge by basic edge detection algorithm, and calculates the gradient direction of each pixel in the edge by canny operator. Then it gets the gray values of the four adjacent pixels on both sides of each edge point in the gradient direction by linear interpolation method, and calculates their difference values. Finally it uses quadratic curve fitting to fit those difference values and gets the sub-pixel edge location. The experimental verification of this algorithm shows high accuracy.