[Qgis-developer] georeferencer plugin - map follow mode

[Maxim Dubinin]

yep, it is much clearer now, we're basically discussing immediate
review of the results of rectification. This sounds good, though can
hit performance for large rasters.

I agree we need to make this feature switchable. I think you should better
update our svn and ask Magnus to patch trunk again with the whole deal. If you're going to
keep working on this I also suggest we move all our stuff from our svn to
QGIS trunk.

Anyway - great job. Now we need to take care of point prediction and
cross-correlations and it'll be perfect.

-- 
Best regards,
 Maxim

Вы писали 18 февраля 2009 г., 21:07:18:

MM> Hi Maxim,

>> Manuel, I've read provided links twice, but still not following ;)
>> "map follow" discussion in regard to Georeferencer. Can you
>> explain?

MM> The basic idea is to use the GCP-derived transforms to show the same
MM> viewport in the qgis main canvas and in the georefencer canvas.

MM> Now that I have reread Maciek's posting, I am also not sure anymore
MM> he described the same, so sorry if I confused you all :-)

MM> What I have currently done is encapsulating the different parametrised
MM> transform methods (linear, helmert, the gdal based polynomial/tps) and
MM> associated parameter fitting in GeorefTransformer classes - this is mainly
MM> some refactoring to avoid code duplication and special cases for the
MM> different transforms. A GeorefTransformer instances provide methods to
MM> fit parameters to the GCPs, and then can be used to transform
MM> either the main canvas viewport into raster coordinates or vice versa, based
MM> on this parametrisation. I want to add checkboxes which give you the option to
MM> link the georeferencer viewport to the main viewport (and vice versa).

MM> This is very useful if you reference a raster to a data source in
MM> the main canvas (e.g. when referencing a scanned street map against a 
MM> road vector layer), and already have a good approximation of the final
MM> transform. Often you only need to add local refinements to bring the residuals
MM> down (especially if you using local interpolation methods like TPS, or a high
MM> order polynomial), and such linked viewports will aid immensly in finding
MM> correspondencies faster.
MM> The transformer classes are also useful to report residuals for the
MM> choosen parametrisation. 

>> Also, please try to coordinate your patches it with us, as there are some
>> more additions sitting in our georeferener SVN.

MM> Is this the repository at http://svn.gis-lab.info/qgis-gdal-georef/?

MM> I am currently developing against qgis trunk. Should I update
MM> to your repository?  I will unfortunatly have a larger patchset than
MM> anticipated - but given the small codebase, merging wouldn't present a problem
MM> I guess (if/when this feature is up for inclusion).

MM> Hope I clarified the idea a bit.

MM> bye,

MM>         Manuel