Sub-Pixel Edge Detection Algorithm Based On Arctangent Function Fitting

Digital image processing technique is widely used in various fields of life. It based on the technique of image edge detection. This technique is used for image preprocessing for the essential step of image understanding and analysis. With the continuous development of human sciences, the edge detection accuracy in all areas of the image is faced with increasingly high demand. So the edge detection technology rises from the initial whole sub-pixel to pixel level. But how to detect the edge of the image is still not satisfactorily resolved research.The traditional whole-pixel edge detection method uses some differential operators to calculate the edge point location, such as Roberts operator, Sobel operator, Prewitt operator, Laplacian operator, Log operator and Canny operator, etc. Since the initial whole pixel edge detection technology has been unable to meet current demand, sub-pixel edge detection techniques have been proposed. The existing sub-pixel edge detection methods are divided into three categories, fitting, interpolation and moment mothod. Although the sub-pixel edge detection method relative to the pixel level has been improved, it still has some deficiencies on the precision and speed. Thus, this paper presents a sub-pixel edge detection method based on arc tangent function fitting. The general steps are as follows:First, canny operator is the best operator in all of the whole pixel level detection operators, so first using Canny get the whole pixel edge location.Second, let the arc tangent function model as a baseline function for image sub-pixel fitting and get a more accurate image edge location.Third, for the two-dimensional edge detection, we can increase the efficiency of edge detection by using the space moment theory to determine the direction projection of the edge, then two-dimensional edge detection problem can be transformed into a one-dimensional problem.Fourth, we verify the arc tangent function fitting high accuracy and robustness by experiment, the arctangent fitting significantly faster than the other edge detection methods on detection time comparison. It can be kown through the arc tangent function fitting and the existing edge detection method. This method can satisfy the demands of image measurement, which is stable, reliable, high accuracy and strong real-time requirements. It has good practicability.