Data is different depending on which band is requested

[gbriggs]

The following two commands give completely different data results for band 2 (the only band in the first case, and the middle band in the second):

gdal_translate.exe -of ENVI -b 2 -srcwin 27200 6450 501 451  jpip://216.150.195.220/JP2Server/Boulder_Foothills "C:\tmp\test_server_gdal_band2.img"

gdal_translate.exe -of ENVI -srcwin 27200 6450 501 451  jpip://216.150.195.220/JP2Server/Boulder_Foothills "C:\tmp\test_server_gdal_all_bands.img"

If you replace the -b 2 with -b 1, the results appear fine.

[gbriggs]

Replying to gbriggs:

The following two commands give completely different data results for band 2 (the only band in the first case, and the middle band in the second): gdal_translate.exe -of ENVI -b 2 -srcwin 27200 6450 501 451  jpip://216.150.195.220/JP2Server/Boulder_Foothills "C:\tmp\test_server_gdal_band2.img" gdal_translate.exe -of ENVI -srcwin 27200 6450 501 451  jpip://216.150.195.220/JP2Server/Boulder_Foothills "C:\tmp\test_server_gdal_all_bands.img" If you replace the -b 2 with -b 1, the results appear fine.

I suspect this issue may be related to the fact that the region decompressor being utilized by JPIPKAK performs color conversions automatically. But I have not confirmed this.

[warmerdam]

I have had no problem reproducing serious issues with the situation described above.

[warmerdam]

The problem seems to be particular to handling of YCbCr images. Boulder_Foothills uses YCbCr as the underlying color space. If three bands are requested at once things are automatically transformed to RGB by the region decompressor. But if you request just band 2 then you get the "Cb" component which does not look like much.

Note that requesting band 1 was not *really* working - it is just that the Y component (intensity) is similar in appearance to any of the red, green or blue channel results.

I *think* the solution is to ensure that all components are fetched from the JPIP server for YCC color model images. I'm going to try this out.

[warmerdam]

This strategy has been implemented in trunk (r19426) and appears to be working properly.