Method of restoring clear image in unmanned aerial vehicle fuzzy noise image

Abstract

The invention relates to an image restoration technology and particularly relates to a method of restoring a clear image in an unmanned aerial vehicle fuzzy noise image. The method comprises steps: the unmanned aerial vehicle fuzzy image is acquired to construct a first optimization equation; a group sparse domain of a clear image obtained by a dictionary learning method and a group sparse domainof a fuzzy kernel matrix are substituted to the first optimization equation to acquire a second optimization equation; and the second optimization equation is substituted to a separated Bregman iteration (SBI) algorithm to solve the fuzzy kernel matrix, and a restored image is obtained. By using the sparse characteristics of the fuzzy kernel and an original image in the group sparse domain, a joint estimation optimization equation is constructed, the fuzzy kernel is estimated while fuzziness is eliminated, the most accurate solution is obtained through the iteration algorithm, and in a condition of thoroughly saving the image details, the fuzzy image is restored.

Description

technical field [0001] The invention relates to the technical field of image restoration, in particular to a method for restoring clear images under fuzzy and noisy images of unmanned aerial vehicles. Background technique [0002] In 2006, Fergus et al. proposed a method based on a priori model in the space domain, which can effectively estimate the point spread function (Foint Fpread Function, PSF) and remove complex camera shake and blur. However, only RL deconvolution is used to reconstruct the image, and the recovery result is The ringing effect is noticeable. In 2007, Jia Jiaya proposed a new method for estimating PSF with image transparency map. In 2008, Shan proposed a blind restoration method that performs PSF estimation and image restoration at the same time, and the effect is good. However, Joshi, Cho and Lee et al. adopted a simplified Gaussian prior model to estimate the blur kernel based on the prediction of the clear boundary of the image, which can achieve a...

AI Technical Summary

Problems solved by technology

[0002] In 2006, Fergus et al. proposed a method based on a priori model in the space domain, which can effectively estimate the point spread function (Foint Fpread Function, PSF) and remove complex camera shake and blur. However, only RL deconvolution is used to reconstruct the image, and the recovery result is Significant ringing effect

However, in the actual restoration, it is impossible to know so much knowledge about the image degradation model in advance, so the restored image effect obtained by this algorithm is not satisfactory.

Images