Publications by authors named "Yeon-Kug Moon"
Article Synopsis
- Recent advancements in super resolution (SR) have utilized generative adversarial networks (GANs) to create more realistic images from low-resolution inputs, but challenges arise from loss of high-frequency details.
- The paper introduces a novel approach using two-dimensional structure consistency (TSC) to analyze image details, along with a mutual loss mechanism that adjusts the training based on TSC filtering.
- This method reduces distortions in high-frequency regions while maintaining clarity in low-frequency areas, outperforming other SR techniques in both qualitative and quantitative evaluations, particularly in perceptual quality metrics like LPIPS.
Sound rendering is the process of determining the sound propagation path from an audio source to a listener and generating 3D sound based on it. This task demands complex calculations, including trigonometric functions. This paper presents hardware-based inverse cosine function calculations using the table method and linear approximation.
View Article and Find Full Text PDF