http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
이미지 초해상도를 위한 SRCNN 기반 데이터 증강 기법
장선규(Seon-Gyu Jang),윤대열(Dai-Yeol Yun),황치곤(Chi-Gon Hwang),이수욱(Soo-Wook Lee) 한국정보통신학회 2024 한국정보통신학회논문지 Vol.28 No.2
Research on display devices capable of implementing high-resolution images is actively underway, leading to increased interest in the production of accessible high-resolution image content. Traditional interpolation algorithms such as Bilinear and Bicubic Interpolation, however, result in a loss of sharpness at the edges of objects, leading to a blurring effect. To address these issues, we employ the Super-Resolution Convolutional Neural Network (SRCNN), which enhances sharpness and is well-suited for generating high-resolution images. As part of our SRCNN-based image data augmentation techniques, we introduce Affine Transformations and Noise Injection methods, specifically Salt and Pepper Noise. These data augmentation methods aim to train the model on diverse image data, incorporating variations in angles and sizes, and increasing data diversity by introducing white or black pixels to the images. We advocate for an improved dataset training approach, enhancing the generalization capability of the existing SRCNN learning model.