[Import XYZ product] performance issue with scaled reflectance raster

[janzandr]

Discussed with @jakimowb and @danschef .

Imported sensor products store proper data gain/offset values for scaling the reflectance data into the 0 to 1 range. When visualizing such raster in QGIS/EnMAP-Box, very large default histograms are stored in the raster with 10 mio bins. This is slowing down raster handling/visualization.

This is most likely a GDAL related bug, which should be reported. As a wordaround, we can just precalculate the histograms with reasonable bin counts and store into the raster.

for band_index in range(1, dataset.RasterCount + 1):
    band = dataset.GetRasterBand(band_index)
    min_val, max_val = band.ComputeRasterMinMax()
    histogram = band.GetHistogram(min=min_val, max=max_val, buckets=1000, approx_ok=False)
    band.SetDefaultHistogram(min_val, max_val, histogram,)

• [ ] procalculate histograms
• [x] report issue to QGIS project

[jakimowb]

Is this caused by gdal or QGIS?

[janzandr]

Visualizing the raster in QGIS will trigger the creation of the histograms.

[janzandr]

I will report it in the QGIS issue tracker. @rouault will hopefully take care of it ;-)

[rouault]

@rouault will hopefully take care of it ;-)

I'm like everybody: I only work 25 hours / day... More seriously: the whole point of open source is that you can try fixing bug that annoy you, without being depend on the availability/willingness of someone else to do it for you.

[janzandr]

Hi @rouault, we will fix this on our side (Python plugin) and hopefully, someone on the C++ side will also take care of it. We here, very value your past effords regarding improving raster integration in QGIS. Have a nice day, looking forward seeing you at one of the next QGIS conferences. All the best, Andreas.