This PR adds tools for correcting global shading on large FIBSEM volumes (e.g., here). These tools are:
Quadratic and fourth order models that fit the values of a function [(x, y) -> v], where x, y are the pixel coordinates (scaled to [-1, 1] x [-1, 1]) and v is the intensity value.
A small ImageJ plugin that allows to choose ROIs and fit such models (using only pixel values in the ROIs) interactively.
A spark client that corrects the intensity of large datasets block-wise.
For supplying model parameters to the spark client, I used an ad hoc json representation, since the model is expected to vary in z. This representation is documented in BackgroundCorrectionClient. Feedback about the documentation and the representation itself is very welcome.
One particular problem with doing this at the level of a stack is that we would need an exported mask to not also correct regions where there is no tissue. Is it safe to assume that a mask will always be exported when background correction is done? In this case, I will add support for masks.
This PR adds tools for correcting global shading on large FIBSEM volumes (e.g., here). These tools are:
For supplying model parameters to the spark client, I used an ad hoc json representation, since the model is expected to vary in z. This representation is documented in
BackgroundCorrectionClient. Feedback about the documentation and the representation itself is very welcome.One particular problem with doing this at the level of a stack is that we would need an exported mask to not also correct regions where there is no tissue. Is it safe to assume that a mask will always be exported when background correction is done? In this case, I will add support for masks.