When rendering data on clients such as OpenLayers, imageProcessing operations are performed to fit the requested windows with data coming from the underlying datasets. You may think about having a huge dataset with a resolution of 50cm and the current rendering is asking for data at 1 km. When overviews have been properly computed as part of preprocessing before configuring the layers, the rendering results are visually good.
When overviews are unavailable, the default logic does a huge subsampling (skip pixels when reading data from disk) before applying the final scaling. The quality of the output isn't good in that case, the rendering is very pixellated.
We should expose the imageMosaic DecimationPolicy parameter in the configuration layer as well as add some more values to that enum and update the decimation code accordingly.
This will allow to do less aggressive subsampling in order to get better quality results.
|