The reconstruction of motion-blurred images and the acquisition of target velocity are widely used in image processing, computer vision and so on. Using traditional optical methods such as Hough transform or differential multiplier, it is difficult to obtain high-resolution and high-definition image quality reconstruction results. Although using deep learning methods for image reconstruction can obtain better results, it will also bring some problems, such as huge data volume, long training time and complex data set construction. In this study, we firstly propose a method based on optimized Radon transform and optical model, combined with adaptive erosion operation, which can process motion-blurred images of various styles and degrees by giving their blur information exactly. Secondly, we construct a sharpness evaluation function scoring algorithm to filter the reconstructed images, which can use blurred information to reconstruct high-fidelity images. Finally, we design a velocity measurement method based on an optical imaging camera system by using blurred prior information. The results show that the blurry image reconstruction and velocity measurement have high robustness and can deal with motion blur in simulations and experiments. The method proposed in this paper can reconstruct a variety of motion-blurred images with high fidelity, and can accurately calculate the motion velocity of the target, which has great application value in the field of image processing, computer vision, and so on.