Multi-Level Generative Chaotic Recurrent Network for Image Inpainting

Cong Chen, Amos Abbott, Daniel Stilwell; Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2021, pp. 3626-3635

Abstract


This paper presents a novel multi-level generative chaotic Recurrent Neural Network (RNN) for image inpainting. This technique utilizes a general framework with multiple chaotic RNN that makes learning the image prior from a single corrupted image more robust and efficient. The proposed network utilizes a randomly-initialized process for parameterization, along with a unique quad-directional encoder structure, chaotic state transition, and adaptive importance for multi-level RNN updating. The efficacy of the approach has been validated through multiple experiments. In spite of a much lower computational load, quantitative comparisons reveal that the proposed approach exceeds the performance of several image restoration benchmarks.

Related Material


[pdf]
[bibtex]
@InProceedings{Chen_2021_WACV, author = {Chen, Cong and Abbott, Amos and Stilwell, Daniel}, title = {Multi-Level Generative Chaotic Recurrent Network for Image Inpainting}, booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)}, month = {January}, year = {2021}, pages = {3626-3635} }