https://bugzilla.wikimedia.org/show_bug.cgi?id=52045
--- Comment #22 from Bawolff (Brian Wolff) <[email protected]> --- (In reply to Bawolff (Brian Wolff) from comment #21) > Experimentation locally suggests that running things through > > vips im_clip2fmt tempInput.v finalOutput.png 2 > > may fix the issue mentioned in comment 2 > > (For reference, the 2 means convert to IM_BANDFMT_USHORT. Previously the > convolution had converted it to IM_BANDFMT_FLOAT, which seems to sometimes > cause problems when outputting as png. I'm unclear as to the why.) Except that breaks other images. What appears to be happening (I think, behaviour locally and on cluster is kind of different, but related), is that image gets read into vips, using a form of either 8bit integer samples, or 16bit integer sample. Then im_convf converts it to floats to do sharpening, and then when output, it gets converted back to integer samples, but incorrectly. What I think we need to do: run vips im_header_int format temp.v before the convolution, save the result, then use im_clip2fmt to convert it back after the convolution. -- You are receiving this mail because: You are the assignee for the bug. You are on the CC list for the bug. _______________________________________________ Wikibugs-l mailing list [email protected] https://lists.wikimedia.org/mailman/listinfo/wikibugs-l
