Opened 10 years ago
Closed 10 years ago
#5640 closed defect (fixed)
Wrong behaviour of average overview with alpha band
Reported by: | Even Rouault | Owned by: | Even Rouault |
---|---|---|---|
Priority: | normal | Milestone: | 1.11.1 |
Component: | GDAL_Raster | Version: | unspecified |
Severity: | normal | Keywords: | overview alpha average |
Cc: |
Description
Currently, if one computes an averaged overview of a RGBA dataset where there's a transition between valid data (alpha=255) and invalid data (alpha=0), the RGB components of the overview will - correctly - ignore the source pixels that have alpha=0. However the alpha values will be averaged. So the end result will be a non-premultiplied alpha RGBA overview, which is unwanted. A premultipled alpha RGBA would perhaps be better, but another acceptable solution, at least in the case where there are only alpha=0 and alpha=255, is to mask the alpha band with itself...
Change History (1)
comment:1 by , 10 years ago
Milestone: | → 1.11.1 |
---|---|
Resolution: | → fixed |
Status: | new → closed |
Note:
See TracTickets
for help on using tickets.
trunk r27655, r27656 , branches/1.11 r27657 : "Overview: ignore alpha=0 values when compute an average overview of an alpha band; and also avoid memory errors when calling GetMaskBand?()/GetMaskFlags() after overview computation if GetMaskXXX() has been called before (#5640)"