<![CDATA[A This study aims to analyze and compare the performance of two artificial intelligence–based image restoration methods, namely GFPGAN (Generative Facial Prior GAN) and RestoreFormer, in enhancing the quality of old photographs of TGKH. Muhammad Zainuddin Abdul Madjid. Both methods were tested on images that had undergone visual degradation, with OpenCV used as a conventional baseline. The evaluation was conducted using two main parameters: processing time and result quality, measured through the Peak Signal-to-Noise Ratio (PSNR) metric. The results show that GFPGAN produced the best outcome, achieving the highest PSNR value (32.8 dB) and the fastest processing time (9 seconds), generating sharp and realistic facial details. RestoreFormer yielded nearly comparable quality with a PSNR of approximately 30 dB and a slightly longer processing time (10 seconds), but it demonstrated greater consistency in preserving the authenticity of textures and structural details. Meanwhile, the manual OpenCV-based method achieved only moderate improvement (25.2 dB) with the longest processing time (12.5 seconds). These findings indicate that AI-based technologies, particularly GFPGAN and RestoreFormer, hold great potential for the preservation of historical visual archives, as they can significantly enhance the quality of old photographs while maintaining their original visual character. The combination of both methods is recommended for digital restoration efforts that balance technical efficiency and cultural authenticity.]]>
Copyrights © 2026
