An identification method for uniform linear motion blurring direction based on second-order difference spectrum was proposed. The power spectrum of the blurred image was calculated after the Laplace second-order difference, and then the power spectrum was processed by the homomorphic filtering and the circular low-pass filtering. The blurring direction was attained by linear fitting of the frequency points with highest amplitude selected from the spectrum image. The experiments were carried out by using the blurred images which were simulated by the Lenna standard images with the blur extent being 20 pixels, and the mean square error of detection were 1.32° without additional noise and 2.27° with Gauss noise that the variance was 0.01. When using the real blurred images, the accuracy of the detection was -0.54°. This proposed method is proved to be available for motion blur of random plane direction and adaptable for noise.