The speckle images of rough surface formed by laser irradiation of workpiece are analyzed using an autocorrelation function to obtain two- order statistical properties. Some parameters of the autocorrelation function are extracted from the speckle images using a fractal algorithm, and a sample set between the fractal dimension and the surface roughness is established and fitted by the least squares polynomial method. The experimental results show that the proposed surface roughness measurement method based on the fractal dimension of laser speckle is feasible and suitable for online high precision roughness detection. The detection time is shortened from tens of seconds to seconds and the detection accuracy reaches to micrometer scale.