Image deblurring method based on multi-task CNN

Abstract

The invention discloses an image deblurring method based on a multi-task CNN. The image deblurring method comprises the following steps: (1) acquiring a total training set, a test set and preprocessing; (2) performing multi-scale zooming and parameter setting on the image; and (3) deblurring based on a multi-task convolutional neural network, wherein the multi-task convolutional neural network comprises three scales, and each scale is divided into an image deblurring module, an image detail recovery module and a feature fusion module; the image deblurring module comprises an image encoding block E and an image decoding block D, and the image encoding block E extracts image features and encodes the image features, and then a deblurred image is obtained through the image decoding block D; the image detail recovery module only comprises a convolution feature extraction layer with a residual structure, and the size of a network feature map is kept the same as the input size, and only blurred image high-order features are extracted and fused, and high-frequency information is provided for the final image recovery process; and the feature fusion module comprises an image merging module and a convolution layer.

Description

technical field [0001] The invention belongs to the field of computer image processing, and is mostly used in related fields such as image or video deblurring, and specifically relates to an image deblurring method based on multi-task CNN. Background technique [0002] During the image acquisition process, due to the shaking of the camera equipment or the rapid movement of the shooting scene during the exposure time, coupled with the influence of atmospheric light, the image quality will be degraded. The degradation of image quality will seriously affect the subsequent processing of images, such as image comparison, feature extraction, image recognition, etc. Due to its cutting-edge and wide application, image deblurring has always been a hot spot in the field of computer vision and image processing. [0003] Existing image deblurring methods can be divided into traditional methods and deep learning-based methods. Traditional methods can be divided into deblurring algorith...

AI Technical Summary

Problems solved by technology

However, at present, most of the convolutional neural networks are used to restore images, and the restoration of image detail information is not considered, resulting in the loss of detailed texture after image deblurring.

On the other hand, the generalization ability and adaptability of single-task network models are not enough, and it is difficult to achieve multi-type and multi-scene image deblurring tasks.

Images