WMS raster transparency problem

[jctull]

It appears that a WMS raster layer with transparency set in the layer properties will increase transparency with every screen refresh until it becomes invisible. To replicate, try the NED shaded relief layer from:

​http://toposervices.cr.usgs.gov/wmsconnector/com.esri.wms.Esrimap/USGS_EDNA_geo?WMTVER=1.1.1&LAYERS=NED_2003_SHADEDRELIEF&SRS=EPSG:4326&FORMAT=PNG&BGCOLOR=0xFF0000&TRANSPARENT=true&SERVICE=WMS&STYLES=&REQUEST=capabilities

[jctull]

I realize now that the link can be shortened to:

​http://toposervices.cr.usgs.gov/wmsconnector/com.esri.wms.Esrimap/USGS_EDNA_geo?

[ersts]

This is a known issue. If you keep repeatedly hitting apply, the transparency it is additive. Likewise, if you decrease the transparency, it is still additive. However, if you pan the map canvas a little and cause a new fetch request, your layer will be at the correct/expected transparency level.

It seems to be a bug in Qt with setting the alpha value for an image more than once. Each apply causes a refresh and thus a reapplication of the alpha value. Too speed things up, the WMS provider does not request a new version of the image if the extent has not changed, it uses a cached version.

I am going to downgrade this a little, while it needs to get fixed but it does not cause a crash or a damage of data

[jctull]

Peter: Thanks for the detailed explanation. I did not see a bug filed, so added it this weekend. It sounds like I need to be filing bugs against Qt instead... Also, it's good to know that forcing a wms refresh works around the problem.

[ersts]

I think it is more just that setAlphaChannel() does not work quite a expected. When you use that method it changes the image format to QImage::Format_ARGB32_Premultiplied

The image is stored using a premultiplied 32-bit [[BR]]
ARGB format (0xAARRGGBB), i.e. the red, green,[[BR]]
and blue channels are multiplied by the alpha component [[BR]]
divided by 255. (If RR, GG, or BB has a[[BR]]
higher value than the alpha channel, the results [[BR]]
are undefined.) Certain operations (such as [[BR]]
image composition using alpha blending) are faster [[BR]]
using premultiplied ARGB32 than with plain ARGB32.

So I changed it around a little have had the WMS Provider store a QImage::Format_ARGB32 then just adjusted the pixel values on each refresh. WMS images should be small enough to not see any performance hit.

The mods were made to the trunk r:10146. If you can take it for a spin and make sure it works more as expected. If it does I will close the ticket and add it to the Version-1_0 branch for inclusion in the next release.

[jctull]

It does appear to be working now. Great! I closed the ticket.

[(none)]

Milestone Version 1.0.1 deleted