作为胶片电影中最常见的谷粒噪声, 其对应的修复算法虽然很多但却存在过度平滑、复杂度高等诸多缺陷. 因此运用 RGB 通道的相关性, 提出一种基于分区域双边滤波的噪声修复算法, 目标是去除胶片电影中常见的谷粒噪声, 为后续斑点、划痕、稳像、闪烁等修复工作提供质量保证. 测试结果表明: 运用本算法去噪后的图像不仅能满足人眼的视觉要求, 而且其峰值信噪比(peak signal to noise ratio, PSNR)值和结构相似性(structural similarity, SSIM)值明显提高, 运行时间也几乎缩短为原来的一半, 这不仅会提高旧电影的商业价值, 而且对挽救国内外大量史实资料具有重要意义.

Grain noise in film films and their corresponding repair algorithms are numerous, but they usually have defects, such as smooth transition and high complexity. Combined with the correlation of RGB channels, a noise restoration algorithm based on a sub-regional bilateral filter was developed in this study to eliminate common grain noise in film films and ensure quality assurance for subsequent repair work, such as spots, scratches, image stabilisation, and scintillation. The test results showed that the de-noised image captured using the proposed algorithm not only satisfied the visual requirements of the human eye but also significantly improved the peak signal-to-noise ratio and structural similarity index measure values and decreased the running time to almost half of the original time. These benefits will not only improve the commercial value of old films but will also save a large amount of historical data.