http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
GAN 기반 공간 적응적 비 정규화를 통한 단일 영상 초해상도 복원
윤종수(Jongsu Yoon),김태현(Taehyeon Kim),최윤식(Yoonsik Choe) 대한전기학회 2021 전기학회논문지 Vol.70 No.2
Despite recent advances in technologies on single image super-resolution using deep neural networks, the key question still remains how to recover finer textures and edges. To solve this super-resolution problem, many recent researches have been using conditional generative adversarial network. However, restoring high resolution images using conditional generative adversarial network is disadvantageous in expressing fine textures and edges because there occurs the loss of spatial and high frequency informations. In this paper, informations on images in different scales are added hierarchically by using a spatially adaptive de-normalization method. This method can restore fine textures and edges of an image by inserting different scale informations for each layers in pyramid structure. In experimental results, the efficiency of the proposed method is proved by showing better performance to restore textures and edges in high quality, comparing with other state-of-the art techniques.
곤돌라형 외벽 유지보수 로봇의 수직위치 센서 개발에 관한 연구
윤종수(Jongsu Yoon),김동엽(Dong Yeop Kim),박창우(Chang-Woo Park) Korean Society for Precision Engineering 2013 한국정밀공학회지 Vol.30 No.4
Demand for high-rising building has arisen. However, its maintenance is usually executed by labour. It could have a severe problem. We proposed a gondola robot to solve it. In this paper, we designed a height estimation sensor for this gondola. It is consist of pan-tilt unit, ARS sensor, and laser sensor. The pan-tilt unit keeps the laser sensor to indicate the gravity direction by referencing the ARS. The laser sensor’s range is vertical distance from gondola to ground. However, if there is an obstacle under the gondola, the distance includes its height. To filter it out, we apply a Kalman filter for the height estimation. If the estimated height is changed extremely, the filter decides that there is an obstacle. Then, it remembers the height of obstacle. Other extreme changes of height estimations are reflected. The experimental results using the proposed sensor system show detail flow of the height estimation.
컨볼루션 희소 코딩을 사용한 Retinex 기반 반사율 분해
윤종수(Jongsu Yoon),최윤식(Yoonsik Choe) 대한전기학회 2020 전기학회논문지 Vol.69 No.3
Color constancy is a feature of the human visual system, which detects the relative inconsistency of the perceived color of an object so that the perceived color may not be changed under various illumination conditions, even when the conditions for observing color are changed. Thus, the Retinex theory was designed in consideration of this color constancy. The physics based Retinex algorithms have been popularly used to effectively decompose the illumination and reflectance of the object. However, if there are many detail areas in the image or the illumination changes rapidly, the illumination and reflectance may not be decomposed properly, because of the violation of the smoothness constraint on illumination. In this paper, we use the convolutional sparse coding model to represent the reflectance in more detail. This allows the reflectance component to provide improved visual quality over conventional methods, as shown in experimental results. Consequently, we can decompose Retinex based illumination and reflectance more precisely, then, reduce the perception gap between humans and machines.