Abstract: As a vehicle license plate identification is only found in the event of an accident or over speed involving key vehicle key accident escape. However, the instantaneous overload of the vehicle captured by the surveillance camera is often blurred due to the rapid movement, which is even unrecognizable by the person. Observation of the image of the plaque is generally a low resolution, suffered serious loss of the surrounding edge information, constitute a blinding to blur the existing method of a huge challenge. The resulting fuzzy motion blur can be viewed as a uniform linear convolution and parametric angle and length. In this paper, we present a new scheme based on the number of deficiencies to determine the core ambiguity. By analyzing several coefficients representing the restored image, we judge that the core angle corresponds to the angle of the real movement of the core when the image based on the restoration is not representative. Then, the core length of the Radon transform of the Fourier transform is estimated. Our program can handle fine motion blur, even though the board is human unrecognizable. We evaluate our focus on real-world images and solve the blind image with various algorithms. The experimental results show that the advantages of our proposed method are robust and robust.
Keywords: Blur, Kernel, license plate, restoration, vehicle.