Image Dehazing Networks Based on Residual Blocks and Feature Fusion

Authors

  • Changxiu Dai

Abstract

Nowadays, the quality of haze images has been greatly improved after defogging, but there are still some problems such as incomplete defogging and color bias. To solve these problems, an image-defogging network based on residual parameter block and feature fusion is proposed in this paper. The network includes the feature module of the extracted image, the color feature and content feature fusion module of the image, and the restored image module. In order to extract image color features and content features effectively, residual blocks and mixed attention are used in the color feature and content feature extraction module, which can remove fog and better maintain the color and content of the original image. Finally, in the image
restoration module, the fused image feature map is non-linear mapped to obtain the image after removing haze. Compared with the existing methods, the proposed method has achieved good image visual effects in both computer-generated scene images and actual scene images.

Keywords: Image dehazing; Feature extraction; Residual block; Feature fusion; Convolutional neural network

Cite As

C. Dai, "Image Dehazing Networks Based on Residual Blocks and Feature Fusion",
Engineering Intelligent Systems, vol. 31 no. 6, pp. 449-456, 2023.




Downloads

Published

2023-11-01

Issue

Vol. 31 No. 6 (2023): International Journal of Engineering Intelligent Systems

Section

Article

License

The submission of a paper implies that, if accepted for publication, it will not be published elsewhere in the same form, in any language, without the prior consent of the publisher. Before publication, authors are requested to assign copyright to CRL Publishing Ltd. This allows CRL to sanction photocopying, and to authorize the reprinting of issues or volumes according to demand. Authors' traditional rights will not be jeopardized by assigning Copyright in this way, as they retain the right to reuse the material following publication, and to veto third-party publication.