A Sub-pixel Edge Detection Method Based On Cubic Spline Interpolation

The edge detection technology is use the image as a carrier of the detection object, it has some advantages, such as:the whole field measurement, high degree of automation and the characteristics of non-contact. With the development of processing technology, both from the advanced science and daily life, it’s need to measure, then the accuracy demand is higher and higher. Therefore, how the technology can be more accurate, faster to detect the target edge become one of the scholar’s research hot topic.The image edge detection algorithms can be divided into the following types:the one is the integer pixel edge detection algorithm; the other is sub-pixel edge detection algorithm. The first method is detect the position of edge, which is by the change of the gradient values. The main algorithms are Sobel operator, Canny operator, Roberts operator and so on. However, these algorithms can only the location in one pixel, can’t satisfied in industrial work. Furthermore, the sub-pixel edge detection algorithm is proposed, it can be better deal with the real-time work, which is on the detection precision, the testing time, and the ability to resist noise.In order to achieve accurately position, combine with the edge of the step form type, This paper puts forward a sub-pixel edge detection method based on cubic spline interpolation, include the following contents:①According to principle of spline interpolation get continuous gray level distribution of one dimension, then realize the position of one dimension.②Using one dimensional curved surface fitting method to obtain the edge projection direction,on the basis of one dimension surface projection direction pixel gray-scale invariant theory, then projecting all three dimension points which is in digital window to the same projection plane, so convert two-dimensional edge detection problem into one-dimension.③Get edge point location and combine with the projection direction to realize the two-dimensional edge positioning.Comparing with the existing sub-pixel edge detection methods, the proposed algorithm has good robustness to noise, and computation time is relatively fast, so it has good applicability in practical application.