Hi,
during the last year a few people complained that icons painted by GeoServer on point symbolizers are somewhat “fuzzy” in their rendering, or too heavily antialiased.
This ticket shows some comparisons and provides a patch to fix it (by making sure we don’t paint the icon at the exact location where it should be painted, but at the closest integer pixel instead): http://jira.codehaus.org/browse/GEOT-4479
Now, I’m wondering what to do about this?
The latest patch proposes a system variable to enable the behavior, however I noticed that while I have a preference for the current behavior, the people that dislike it are really really pissed off by it.
So I’d like to hear from others here, what should we do?
I guess one way could be to roll a config in the WMS panel, but if we have a striking majority of people that likes it better one way or the other, I guess the system variable would still be good to accommodate the minority that likes it the other way (and I’ll be happy to sit in that minority if needs be, my preference is slight, I can live with the other behavior just fine too).
I don’t have a strong opinion about it but my first thought is that while I do like the look of the heavier anti aliased symbols it does look a bit strange when the discrepancy occurs. So i could see the case for enabling the behaviour by default to ensure things look consistent.
But then again options are always good and give those that really have an issue with the current behaviour a way out. It also seems a more conservative (safer) approach to leave the current behaviour as is and have the option.
Hi,
during the last year a few people complained that icons painted by GeoServer on point symbolizers are somewhat “fuzzy” in their rendering, or too heavily antialiased.
This ticket shows some comparisons and provides a patch to fix it (by making sure we don’t paint the icon at the exact location where it should be painted, but at the closest integer pixel instead): http://jira.codehaus.org/browse/GEOT-4479
Now, I’m wondering what to do about this?
The latest patch proposes a system variable to enable the behavior, however I noticed that while I have a preference for the current behavior, the people that dislike it are really really pissed off by it.
So I’d like to hear from others here, what should we do?
I guess one way could be to roll a config in the WMS panel, but if we have a striking majority of people that likes it better one way or the other, I guess the system variable would still be good to accommodate the minority that likes it the other way (and I’ll be happy to sit in that minority if needs be, my preference is slight, I can live with the other behavior just fine too).
I think “snapping” to the closest pixel location is great - and should be the default (we can still perform anti-aliasing in this case, but by aligning to the pixel grid we should have a noticeable improvement in quality).
I was also wondering if there is any advantage in quality for performing the same “snap” for text rendering? Probably only for fonts that are defined with pixels in mind.
–
Jody Garnett
On Sunday, 16 June 2013 at 4:06 AM, Andrea Aime wrote:
Hi,
during the last year a few people complained that icons painted by GeoServer on point symbolizers are somewhat “fuzzy” in their rendering, or too heavily antialiased.
This ticket shows some comparisons and provides a patch to fix it (by making sure we don’t paint the icon at the exact location where it should be painted, but at the closest integer pixel instead): http://jira.codehaus.org/browse/GEOT-4479
Now, I’m wondering what to do about this?
The latest patch proposes a system variable to enable the behavior, however I noticed that while I have a preference for the current behavior, the people that dislike it are really really pissed off by it.
So I’d like to hear from others here, what should we do?
I guess one way could be to roll a config in the WMS panel, but if we have a striking majority of people that likes it better one way or the other, I guess the system variable would still be good to accommodate the minority that likes it the other way (and I’ll be happy to sit in that minority if needs be, my preference is slight, I can live with the other behavior just fine too).
