Take shots of color checker and create LUT based on pairs of tif image to match color of two cameras. I used this to match my Nikon Z6iii to Hasselbald 907x 50c.
python generate.py <base_folder_path> <method_flag> <display_patch>The base folder must contain two subfolders, "nikon" and "hasselblad", each containing calibration images (.tif and/or .TIF).
For method_flag, use one of the following flags (default is --standard):
--standard : Global linear transformation using Colour's matrix_colour_correction.
(Computes a 3x3 matrix mapping Nikon patches to Hasselblad patches.)
--robust : Robust affine transformation using RANSAC (3x4 matrix with intercepts).
(Fits a separate linear model per channel and is robust to outliers.)
--nonlinear : Nonlinear (polynomial) transformation using a degree-2 polynomial.
(Captures nonlinearities in the color mapping.)
--piecewise : Piecewise linear transformation that segments the data by luminance
(shadows, midtones, highlights) and interpolates between separate matrices.
--mlp : Neural network mapping using an MLPRegressor.
(A simple feedforward network learns the mapping.)
--iterative : Iterative refinement of the standard matrix using gradient descent.
(Starts from the standard solution and refines it.)
--kernel : Kernel regression (Nadaraya–Watson) with a Gaussian kernel.
(Predicts outputs as a weighted average of calibration points.)
--loess : LOESS (locally weighted linear regression) for local models.
(Fits a local linear model for each new input using Gaussian weighting.)
--local : Local linear models using clustering (e.g. k-means).
(Clusters the data and fits separate linear models in each cluster.)
--ridge : Regularized regression (Ridge) for a global affine mapping.
(Uses L2 regularization for stability.)
--compare : Compare all methods and view their R² values, exposure gains, and previews.
For display_patch, use the following flag (no flag means display patches):
--no-display-patches : Do not display the color patch images (color checker detection).
Only the final result will be shown.
In my case, LOESS takes a huge amount of time (3 hours in my case), but it is the most accurate.
I used mostly ChatGPT for this, as I have no idea how color works. So feel free to modify anything that is incorrect.