http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
장준보(Junbo Jang),임헌성(Heunseung Lim),이재희(Jaehee Lee),최형기(Hyunggi Choi),백준기(Joonki Paik) 대한전자공학회 2022 대한전자공학회 학술대회 Vol.2022 No.11
Although eliminating ring ghost has been industrially required, it is only removed from data biased to certain phenomena due to lack of learning data, and there are limitations to performance. To solve this problem, this paper proposes a method for removing ring-ghost harmful noise. The proposed method is performed by in-painting for detecting the ring ghost region and removing ring ghost noise, and it can provide quality improvement results by removing ring pattern noise from specific lenses and cameras.
5 방향 그라디언트 손실 함수를 사용한 트랜스포머 기반 단일 영상 비 제거 방법
최진솔(Jinsol Choi),임헌성(Heunseung Lim),최형기(Hyungki Choi),백준기(Joonki Paik) 대한전자공학회 2023 대한전자공학회 학술대회 Vol.2023 No.6
In this paper, we propose a novel approach to single image rain removal by incorporating a transformer structure into a U-shaped neural network and implementing a five-direction loss function. Our approach specifically employs a swin-transformer combined with residual and Fourier convolution methods. Beyond conventional x and y directional gradients, we propose a unique technique to eliminate non-components using a five-direction gradient. Experimental results indicate that our method outperforms existing approaches by minimizing information loss and delivering enhanced quality in rain-free images.