...
- Resampling is a user requirement
- Users want to save their preferred resampling options per product type (or maybe even in general)
- Resampling shall be applicable to multisize products
- (1) SceneTransforms shall be SceneTransform.IDENTITY
- (2) All band sizes must be int-divisible by the scaling factor (and its inverse) ← this requirement is important to enable aggregation for the moment
- (3) Either width and height of the reference band must be both smaller or both bigger than (or both equal to) width and height of the bands to be resampled
- For interpolation, the operator shall simply scale the images, using one of the pre-defined JAI interpolations (allow nearest neighbour, bi-linear, cubic convolution)
- For aggregation, the operator shall first use a filter from ceres-jai (first, mean, median, min, max) to determine which value shall be selected
- For complex cases (cases which do not match (1), (2) or (3)), the ReprojectionOp shall be used
- If a tool requires resampling / reprojection, a popup dialog shall appear, asking the user to choose one of the ops (if available) or to cancel
- In the UI, users shall be enabled to set interpolation methods and aggregation methods which are then to be applied to all bands (in future versions, methods can be set to individial bands distinctively)
- CAREFUL with flags and index codings: Values must not be different from 0 or 1
- Allow setting for flag aggregation and index aggregation (two parameters): as above
- For flag and incex interpolation, always use nearest neighbour
- In later versions, aggregation methods might be stated to individual flag bands and even flags
- In future releases, the ResamplingOp shall also be able to resample back to multi-size products. This can be used to e.g. resample bands back to their original sizes after an operator has been used. To achieve this, it is necessary to assign to each band its own reference size/resolution.
When to use? What is its purpose?
...