摘要
为解决学院色彩编码系统(ACES)流程中黑箱摄影机输入设备转换(IDT)构建的难题,本文针对传统一维查找表(1D LUT)方法在对数变换逆推中精度不足与平滑性差的问题,提出一种优化的IDT自构建方法。该方法的核心是采用ACES通用查找表格式标准中定义的参数化分段函数模型,通过构建带约束的最小二乘优化问题,对摄影机的对数曲线进行逆向拟合。该方法创新性地引入实测黑位值作为暗部锚点,并施加函数可微条件约束,从而解决了传统方法在暗部精度差和断点选择不稳的缺陷,显著提升了对数到线性转换的准确性与平滑度。对基于该方法构建的完整工作流进行拟合误差分析表明,优化后的模型曲线与官方标准曲线高度一致,平均绝对误差(MAE)、均方根误差(RMSE)及可决系数(R2)均达到优良值。实验证明,该工作流生成的IDT与官方IDT在视觉及数据层面高度吻合,为黑箱摄影机高效接入ACES流程提供了有效且可行的技术方案。
To address the challenge of constructing input device transform(IDT)for black box cameras within the Academy Color Encoding System(ACES)workflow,this paper proposes an optimized IDT self⁃construction method that targets the problems of insufficient accuracy and poor smoothness found in traditional one⁃dimensional lookup table(1D LUT)methods used for logarithmic transform inverse⁃fitting.The core of this method is to adopt the parameterized piecewise function model defined in the ACES common LUT format standard,to perform an inverse⁃fitting of the camera's logarithmic curve by constructing a constrained least⁃squares optimization problem.This method innovatively introduces the measured black level value as an anchor point for the shadow region and applies a function differentiability constraint.This approach resolves the deficiencies of traditional methods,such as poor accuracy in shadow areas and unstable break⁃point selection,thereby significantly enhancing the accuracy and smoothness of the logarithmic⁃to⁃linear conversion.A fitting error analysis of the complete workflow constructed based on this method shows that the optimized model curve is highly consistent with the official ACES standard curve,with mean absolute error(MAE),root mean square error(RMSE),and the coefficient of determination(R2)all reaching excellent values.Experiments demonstrate that the IDT generated by this workflow shows a high degree of agreement with the official IDT on both visual and data levels,providing an effective and practical technical solution for the efficient integration of black box cameras into the ACES workflow.
作者
徐理航
顾晓娟
XU Lihang;GU Xiaojuan(School of Intelligent Imaging Engineering,Beijing Film Academy)
出处
《现代电影技术》
2025年第12期60-66,共7页
Advanced Motion Picture Technology
