Abstract: A sub-pixel edge detection algorithm based on Bessel edge model is presented.Firstly,the algorithm import a modifiable parameter to the Bessel point spread function,and convolve with step edge model to obtain the Bessel edge model.Then Least-square curve fitting method is used to estimate the model by making use of the image edge data,simultaneously considering the effect of digitization to curve fitting.Finally,the sub-pixel location of image edge is obtained.Experimental results indicate that the average error of sub-pixel location is 3% of a pixel,and the variance of error is 0.000 5.It can satisfy the system requirements of non-contact,higher precision and stabilization,as well as strong robustness to the image noise.