Nowadays, for data interpolation, invented smoothed linear interpolation is used. It works very fast, but has problems:
Does not work on dense data, data clouds: includes every point;
Does not take into account the uncertainty of the input data;
Cannot predict the confidence interval of the resulting spectrum.
Kriging is perfect for these purposes; more details on Wikipedia. There are only two drawbacks to the practical implementation:
It's slower than my smoothed linear interpolation;
For image processing, I'll have to write it without relying on libraries to make it fast.
This will greatly improve the accuracy of the colors, but will require a lot of effort, so it is postponed until at least the summer of 2024. Fortunately, no architectural changes will be required; the code was designed ready for this interpolation update.
Nowadays, for data interpolation, invented smoothed linear interpolation is used. It works very fast, but has problems:
Kriging is perfect for these purposes; more details on Wikipedia. There are only two drawbacks to the practical implementation:
This will greatly improve the accuracy of the colors, but will require a lot of effort, so it is postponed until at least the summer of 2024. Fortunately, no architectural changes will be required; the code was designed ready for this interpolation update.