• 한국컴퓨터정보학회논문지
• /
• 제29권5호
• /
• pp.21-29
• /
• 2024

CT 촬영 시 방사선량을 줄이면 피폭 위험성을 낮출 수 있으나, 영상 해상도가 크게 저하 될 뿐아니라 잡음(noise) 발생으로 인해 진단의 효용성이 떨어진다. 따라서, CT 영상에서의 잡음제거는 영상복원 분야에 있어 매우 중요하고 필수적인 처리 과정이다. 영상 영역에서 잡음과 원래 신호를 분리하여 잡음만을 제거하는 것은 한계가 있다. 본 논문에서는 웨이블릿 변환 기반 GAN 모델 즉, WT-GAN(wavelet transform-based GAN) 모델을 이용하여 CT 영상에서 효과적으로 잡음 제거하고자 한다. 여기서 사용된 GAN 모델은 U-Net 구조의 생성자와 PatchGAN 구조의 판별자를 통해 잡음제거 영상을 생성한다. 본 논문에서 제안된 WT-GAN 모델의 성능 평가를 위해 다양한 잡음, 즉, 가우시안 잡음(Gaussian noise), 포아송 잡음 (Poisson noise) 그리고 스펙클 잡음 (speckle noise)에 의해 훼손된 CT 영상을 대상으로 실험하였다. 성능 실험 결과, WT-GAN 모델은 전통적인 필터 즉, BM3D 필터뿐만 아니라 기존의 딥러닝 모델인 DnCNN, CDAE 모형 그리고 U-Net GAN 모형보다 정성적이고, 정량적인 척도 즉, PSNR (Peak Signal-to-Noise Ratio) 그리고 SSIM (Structural Similarity Index Measure) 면에서 우수한 결과를 보였다.

• Journal of Semiconductor Engineering
• /
• 제2권1호
• /
• pp.125-129
• /
• 2021

As the interest of deep learning, techniques to control the color of images in image processing field are evolving together. However, there is no clear standard for color, and it is not easy to find a way to represent only the color itself like the color-palette. In this paper, we propose a novel color palette extraction system by chroma fine-tuning with reinforcement learning. It helps to recognize the color combination to represent an input image. First, we use RGBY images to create feature maps by transferring the backbone network with well-trained model-weight which is verified at super resolution convolutional neural networks. Second, feature maps are trained to 3 fully connected layers for the color-palette generation with a generative adversarial network (GAN). Third, we use the reinforcement learning method which only changes chroma information of the GAN-output by slightly moving each Y component of YCbCr color gamut of pixel values up and down. The proposed method outperforms existing color palette extraction methods as given the accuracy of 0.9140.

• 한국재료학회지
• /
• 제15권2호
• /
• pp.106-111
• /
• 2005

Effects of stainless steel-making dusts and binder content on compacting $density(\rho)$ and magnetic properties were evaluated. Cores compacted with the mixture of pure Fe powders, $5wt.\%$ dusts and $0.25wt.\%$ binder showed good AC magnetic properties. For example, permeability$({\mu}a)$ and core loss(P) of the cores containing $5wt.\%$ dusts at 500 kHz were 62 and $4008\;{\mu}W/cm^3$, respectively. These properties are almost equivalent to those of competitor's products (i.e, Ancorsteel TC 80 produced by $H\ddot{o}gan\ddot{a}s$ Corp.). The powdered cores obtained from the present work are expected to apply for high-performance soft magnetic components such as normal mode choke filter and pulse transformer.

• Nuclear Engineering and Technology
• /
• 제46권2호
• /
• pp.147-158
• /
• 2014

In order to investigate how the microstructure of fuel/matrix-interaction (FMI) layers change during irradiation, different U-7Mo dispersion fuel plates have been irradiated to high fission density and then characterized using scanning electron microscopy (SEM). Specifially, samples from irradiated U-7Mo dispersion fuel elements with pure Al, Al-2Si and AA4043 (~4.5 wt.%Si) matrices were SEM characterized using polished samples and samples that were prepared with a focused ion beam (FIB). Features not observable for the polished samples could be captured in SEM images taken of the FIB samples. For the Al matrix sample, a relatively large FMI layer develops, with enrichment of Xe at the FMI layer/Al matrix interface and evidence of debonding. Overall, a significant penetration of Si from the FMI layer into the U-7Mo fuel was observed for samples with Si in the Al matrix, which resulted in a change of the size (larger) and shape (round) of the fission gas bubbles. Additionally, solid fission product phases were observed to nucleate and grow within these bubbles. These changes in the localized regions of the microstructure of the U-7Mo may contribute to changes observed in the macroscopic swelling of fuel plates with Al-Si matrices.