基於多尺度融合機制去除摩爾紋的網路模型

Image Demoireing using Multi-scale Fusion Networks

侯治祥(Hou Chih-Hsiang)

指導教授 ： 彭彥璁

國立政治大學/資訊學院/資訊科學系/碩士(2022年)

https://doi.org/10.6814/NCCU202201419

摘要

因為被拍攝的螢幕的顯示器的像素排列，與手機像素的排列出現干涉現象，兩個排列在疊加的過程中就會形成出現了彩色和形狀不規律的條紋，就是摩爾紋。與其他影像還原的任務不同的是，去除摩爾紋的困難點在於，摩爾紋出現的頻率域很廣，不只存在於高頻，也同時出現在低頻中。此外，摩爾紋的形狀是不規則的，摩爾紋的色彩也會產生扭曲，所以是一個有挑戰性的任務。本論文提出一個基於多尺度融合機制去除摩爾紋的網路模型和利用摩爾紋的轉移做資料擴增的方法，可以增強去摩爾文的表現，根據實驗的結果顯示，我們的模型比去摩爾紋領域方法表現的更好。

關鍵字

影像處理 ； 影像還原 ； 影像去除摩爾紋

並列摘要

Taking images on a digital display may cause a visually annoying optical effect, called moiré, which degrades image visual quality. Because the pixel arrangement of the display of the screen being photographed interferes with the pixel arrangement of the phone, the two arrangements are superimposed in the process of forming the color and shape irregularities of the stripes, which are moire patterns. Unlike other image restoration tasks, the difficulty in removing moire patterns is that moire patterns appear in a wide range of frequencies with irregular shapes and rainbow-like colors. Thus, removing moiré patterns is a challenging task. In this thesis, we propose an Image Demoiréing Multi-scale Fusion network (DMSFN) to remove Moiré patterns and a method for data augmentation using the transfer of Moiré patterns, which can enhance the performance of demoiréing. According to the experimental results, our model performs favorably against state-of-the-art demoiréing methods on benchmark datasets.

並列關鍵字

Image processing ； Image restoration ； Image moiré removal

參考文獻

[1] Yujing Sun, Yizhou Yu, and Wenping Wang. Moiré photo restoration using multiresolution convolutional neural networks. IEEE Transactions on Image Processing, 27(8):4160–4172, 2018.

[2] Shanghui Yang, Yajing Lei, Shuangyu Xiong, and Wei Wang. High resolution demoire network. In 2020 IEEE International Conference on Image Processing (ICIP), pages 888–892. IEEE, 2020.

[3] Bin He, Ce Wang, Boxin Shi, and Ling-Yu Duan. Mop moire patterns using mopnet. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 2424–2432, 2019.

[4] Xi Cheng, Zhenyong Fu, and Jian Yang. Multi-scale dynamic feature encoding network for image demoiréing. In 2019 IEEE/CVF International Conference on Computer Vision Workshop (ICCVW), pages 3486–3493. IEEE, 2019.

[5] Xiaotong Luo, Jiangtao Zhang, Ming Hong, Yanyun Qu, Yuan Xie, and Cuihua Li. Deep wavelet network with domain adaptation for single image demoireing. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, pages 420–421, 2020.