人工智能(AI)辅助下的胶片画质增强方法研究OA

To address the problems of film grain loss and image over-sharpening commonly found in existing general artifi-cial intelligence(AI)restoration models when processing archival films,this paper explores a restoration solution that bal-ances sharpness enhancement with texture preservation,taking A Girl from Hunan as a research subject.We propose an AI-assisted enhancement method based on frequency separation and re-fusion.Initially,the film grain is separated from the image content;the content is then enhanced via a dual-path enhancement module and a frequency fusion module.Subsequently,residual superposition synthesis and simulated grain generation are introduced to mitigate issues such as grain loss,over-smoothing,and sharpening distortion inherent in traditional super-resolution algorithms.Subjective evalua-tion results indicate that this method achieves the highest comprehensive preference score compared to the original foot-age and versions using only general AI super-resolution models.This method effectively enhances the resolution of digital film copies while successfully preserving the unique grain structure and aesthetic style of the original physical film medium,thereby achieving a balance between technical restoration and artistic reconstruction.