Bi-Level optimization method for image deblurring

Abstract

The present invention relates to Bi-Level optimization method for image deblurringso that the two levels of loss functions alternate model optimization, Bi- Level optimization mechanism is divided into two steps, the first step is to train a basic model with MSE loss condition, the second step is to use two-level loss iteration to fine-tune the model. This is because at the beginning of training,the divergence between the restoration effect and the clear image is large, and the effect of noise can be ignored, but at the end of training, the noise is amplified and the negative effect is more obvious, so the perceptual loss is introduced to suppress the noise and the MSE loss is changed to L1 loss to ensure the structural continuity sufficiently. The invention has the advantages of accuratetexture restoration and matching depth features with pixel values.

Description

technical field [0001] The invention relates to the field of digital image processing, in particular to a Bi-Level optimization method for image deblurring, which proposes a Bi-Level optimization method in the blurred image restoration process. technical background [0002] Deblurring techniques are a widely studied topic in image and video processing. The blur caused by camera shake seriously affects the imaging quality and visual perception of the image in a certain sense. As an important branch of image preprocessing, the improvement of deblurring technology directly affects the performance of other computer vision algorithms, such as foreground segmentation, object detection, behavior analysis, etc.; at the same time, it also affects the encoding performance of images. Therefore, it is imperative to study a high-performance deblurring algorithm. [0003] Documents 1-3 are the background information of the deep learning deblurring algorithm compared in the present inven...

AI Technical Summary

Problems solved by technology

[0007] The purpose of the present invention is to: propose a Bi-Level optimization method for image deblurring, the Bi-Level optimization method optimizes GAN (Generative Adversarial Network), aiming to solve the shortcomings of existing deep learning deblurring algorithms

[0009] The MSE loss can guarantee the identity of the optimization process at the pixel level and the feature level, but the problem is that it introduces a lot of noise; while the perceptual loss is a good alternative to a certain extent, but it cannot guarantee the optimization. identity

Images