[OSM-talk] OSM gets mentioned in KDE4 keynote speech at Google Headquarters

Artem Pavlenko artem.mapnik at googlemail.com
Wed Jan 23 13:33:50 GMT 2008


On 23 Jan 2008, at 12:23, Torsten Rahn wrote:

> Hi Artem,
>
> On Wednesday 23 January 2008 11:06:41 Artem Pavlenko wrote:
>> Hi Torsten,
>> Glad you like the screenshot. I'll post another one soon :)
>> Yes, you're right  mercator would be wasteful. When I got spare
>> moment I'll try different tile sizes to see if this affects final map
>> quality.
>
> No, the tilesize shouldn't affect the render quality at all.
Ok.
>
> Manually adjusting the point at which Marble switches to a  
> different tile
> level could have a positive effect.
>
> Right now it happens in
>
> void AbstractScanlineTextureMapper::selectTileLevel( ViewParams*  
> viewParams )
>
> using the formula
>
> double  linearLevel = ( 2.0 * (double)( radius )
> 			       / (double) ( m_tileLoader->tileWidth() ) );
>
> I guess ideally the size of the tile on the screen would match  
> about the size
> of the original tile. You can check this by changing line 119 in
> marble/src/lib/TextureTile.cpp from
>
>   bool tileIdVisible = false; if(tileIdVisible)  
> m_painter.paintTileId(theme);
>
> to
>
>   bool tileIdVisible = true; if(tileIdVisible) m_painter.paintTileId 
> (theme);
>
>
> This will enable you to see the tile borders and the filenames +  
> tile level
> visually directly on the screen. This "Debug Mode" is very helpful  
> when you
> work with tiles in Marble!

Thanks, I'll try this.
>
>> Maybe you can afford better resampling at high zooms for smoother
>> results?
>
> I'm not sure whether this can be implemented easily at sufficient  
> speed. Right
> now Marble just looks up the "exact" color value on the map for  
> each point on
> the screen
> We could try to do some more sophisticated high quality rendering  
> when the
> user stops moving the map.

I think this could work. if you're planning to pull tiles from a web  
service, then rendering fast (scaled) globe and repainting on each  
tile arrival (with bilinear interpolation for example) shouldn't be a  
problem. I did something like that - pulling tiles from VE and  
warping them on-fly into NatGrid and it worked really well (qt4).

> Whether it's possible to do the same at acceptable
> speed while moving is a completely different question though ....

I think for 'moving' current approach works well.
>
>>   BTW, I rendered tiles (for marble) in Lat/Lon aka Geographic, Plate
>> Carre, equirectangular,cylindrical equidistant, unprojected etc.
>> Still, there are too many re-sampling artifacts. I'll try smaller
>> tile sizes ..
>
> Could you send me a screenshot of Plate Carre, so I get the idea  
> how it looks
> like?
>

Here you go : http://artem.dev.openstreetmap.org/files/marble-osm- 
flat.jpg
I found 'magic' zoom level when there's minimal distortion.

Cheers
Artem
>
> Best Regards,
> Torsten
>
> PS: Is the OSM-Talk Moderator on holidays?
>
>>> Best Wishes,
>>>
>>> Torsten
>>>
>>>> Best
>>>> Artem
>>>>
>>>>> Best wishes,
>>>>>
>>>>> Torsten
>>>>>
>>>>> --
>>>>>  Torsten Rahn
>>>>>
>>>>>  Tel.: 0 21 61 - 46 43 - 192
>>>>>
>>>>> credativ GmbH, HRB Mönchengladbach 12080
>>>>> Hohenzollernstr. 133, 41061 Mönchengladbach
>>>>> Geschäftsführung: Dr. Michael Meskes, Jörg Folz
>>>
>>> --
>>>  Torsten Rahn
>>>
>>>  Tel.: 0 21 61 - 46 43 - 192
>>>
>>> credativ GmbH, HRB Mönchengladbach 12080
>>> Hohenzollernstr. 133, 41061 Mönchengladbach
>>> Geschäftsführung: Dr. Michael Meskes, Jörg Folz
>
>
>
> -- 
>  Torsten Rahn
>
>  Tel.: 0 21 61 - 46 43 - 192
>
> credativ GmbH, HRB Mönchengladbach 12080
> Hohenzollernstr. 133, 41061 Mönchengladbach
> Geschäftsführung: Dr. Michael Meskes, Jörg Folz





More information about the talk mailing list