<![CDATA[The modern broadcasting industry requires visual content in high-definition (HD) resolution. This technological transition creates critical issues in digital asset management, specifically regarding legacy broadcasting archives that are primarily recorded in standard definition (SD) formats. Conventional spatial interpolation approaches fail to address this issue, producing blurry and pixelated images. The application of artificial intelligence, particularly the Real-ESRGAN algorithm, offers a promising restoration solution. However, this implementation significantly increases the raw file size, making it impractical for local server storage. This research aims to find a precise equilibrium between visual quality enhancement using AI and storage efficiency through High-Efficiency Video Coding (HEVC) compression. This study uses a quantitative experimental method via laboratory-scale software engineering. The intervention phases include pixel reconstruction using the AI model and file size reduction using the HEVC standard. Data collection involved ten SD video samples. Real-ESRGAN qualitatively restores texture details and removes analog noise without excessive artifacts. Due to file size expansion, further compression using HEVC with varying Constant Rate Factor (CRF) parameters is being conducted to reduce the size by over 50% without severe Video Multimethod Assessment Fusion (VMAF) metric degradation. The integration of AI upscaling and HEVC compression creates a highly applicable workflow for television industry needs.]]>
Copyrights © 2026
