A Video Motion Blur Method Strengthening Character Priors

Abstract

The invention discloses a video motion blur removal method that strengthens character priori. Firstly, a target tracking algorithm is used to obtain target continuous motion frames, and motion estimation is performed on two adjacent frames of images. Using the relationship between exposure time and frame time, through phase Two adjacent frame motion vectors estimate the initial value of the point spread function; then according to the image and the initial point spread function, the regularization formula proposed by the present invention is used to solve the point spread function, and the point spread function obtained is used to carry out non-blind de-motion blurring of the image, A corresponding clear image is obtained. Aiming at the blurring caused by collecting moving objects when the camera is fixed, the present invention proposes a motion blurring method that strengthens character priori, and improves the restoration ability of motion blurred video.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and more specifically relates to a video motion blur removal method that strengthens character priors. Background technique [0002] Users collect videos for viewing or analysis. Regardless of the purpose, the blur caused by the movement of the target will cause trouble to the user, especially in the analysis of the blur caused by the movement of the target in the video. Here comes the difficulty. [0003] For the video restoration problem, there are generally two solutions at this stage: from the perspective of hardware assistance, design a camera or camera that can prevent target motion blur; from the perspective of algorithms, there are generally two common methods: one is to estimate the blur first The function type and parameters transform the blind defuzzification problem into a non-blind defuzzification problem, and then use the non-blind defuzzification algorithm to deconvolute a...

AI Technical Summary

Problems solved by technology

This method requires the support of some hardware devices, resulting in an increase in cost and relatively low convenience

Yuan uses an image without motion blur but with noise as auxiliary information. This algorithm has a good effect on the blur caused by fog, light and other environments, but in many cases it is impossible to obtain an image with only noise, so this algorithm is practical. limited sex

Images