Blind image deblurring method based on depth prior

Abstract

The invention discloses a blind image deblurring method based on depth prior. A deep convolutional neural network DIP-Net is used for implicitly modeling an image smoothness prior constraint to generate a clear image; estimating a fuzzy kernel by solving an accurate solution about a fuzzy kernel optimization problem; and alternately iteratively updating the blurring kernel and the clear image, calculating a loss function by using the restored clear image and the blurring kernel, and updating network parameters. Carrying out joint modeling on the blurred image and the blurred kernel, and simultaneously estimating a clear image and the blurred kernel by adopting a mode of alternately iterating a network model and a mathematical model; blind deblurring of end-to-end self-supervised learning is achieved only using blurred images without any additional implicit or explicit image priors. The regularization method is realized in combination with a deep network structure, and a blurred image and a blurred kernel truth value training network do not need to be used; compared with a traditional model method, the method does not need to employ an image pyramid mode to estimate the blurring kernel from coarse to fine, and effectively inhibits the noise in the restored image.

Description

technical field

[0001] The present invention relates to the field of image deblurring, and more specifically, relates to a blind image deblurring method based on depth prior. Background technique

[0002] During the image acquisition process, due to the influence of factors such as atmospheric turbulence, relative motion between the imaging device and the target, and inaccurate focusing of the imaging device, the acquired image will be blurred to a certain extent. In many fields such as traffic monitoring, biomedicine, astronomical observation, remote sensing and telemetry, clear images can provide more useful information. In order to meet the needs of clear images in various application fields, we generally start from two aspects of hardware and software. There are problems such as high cost, high technical difficulty, and susceptibility to environmental influences by improving hardware, while image deblurring technology refers to recovering clear images from blurred image...

Images