[Qgis-developer] raster visualization

aperi2007 aperi2007 at gmail.com
Sun Apr 22 15:37:06 EDT 2012


Hi Agus ,
thx for your informations.

 > Opening the raster layers very fast but getting a bad visualization
 > would not be a major
 > waste of time? This is what currently happens to me: the initial
 > display is always useless.
 >

I understand your point of view, but i guess is not a need for every 
users. Same users can't wait the 4+ minutes need for the elaboration of 
a raster.
Again because an user could open usually more than 20 rasters for every 
session, I guess this elaboration should be absolutely optional to avoid 
a really big lost of time.

 > If your raster layers have similar statistics and or you want to
 > display them with the same stretching so
 > that grey levels or colors are comparable (which is often my case
 > also, with time series of ndvi for example), a common style file would
 > be the best solution.

Your solution seem good to me.

We have more set of rasters of ortophoto type colors and grey.
Eachset of colors photo is surely with comparable levels, instead I 
don't know for the gray photo sets.
They are too older (1954 and so on).

Instead for raster cell (grid data) from Lidar:
Just now we begin to merge all our sets to have a full land-cover.
So we are merging 1x1, 2x2 and 3x3 meters cells.
I don't know if this could have some trouble with having a comparable 
set of values, but guess of no.

Surely is not possible to find a unique style good for all the sets, but 
is possible to establish a good style for each set.

 >...Perhaps, in
 > case a raster style is present, the by default procedure and
 > calculations could be bypassed. Would this be a good solution
 > in your case?
 >

You mean a file raster-style in the same folder of the raster set or of 
the catalog file ?
I guess this could be a good compromise to help the use in situation 
like ours.

Andrea.

Il 22/04/2012 19:43, Agustin Lobo ha scritto:
> Andrea,
>
> Opening the raster layers very fast but getting a bad visualization
> would not be a major
> waste of time? This is what currently happens to me: the initial
> display is always useless.
>
> If your raster layers have similar statistics and or you want to
> display them with the same stretching so
> that grey levels or colors are comparable (which is often my case
> also, with time series of ndvi for example), a common style file would
> be the best solution. Perhaps, in
> case a raster style is present, the by default procedure and
> calculations could be bypassed. Would this be a good solution
> in your case?
>
> Agus
>
> El día 20 de abril de 2012 15:10, Andrea Peri<aperi2007 at gmail.com>  escribió:
>>> - on initial load of a raster, generate a quicklook that is the larger
>>> of 1/4 screen resolution or 500x500 pixels by sampling every nth pixel
>>> - generate a histogram from the quicklook
>>> - calculate clipped 2% - 96% range min max for each band
>>> - apply a histogram stretch based on the above
>>> - the histogram could be used for generating the graph in raster
>>> props, and the quicklook could be used to create thumbs and previews
>>> etc
>>> - ideally we should cache these quicklooks and only regenerate them if
>>> the underlying dataset has changed
>>>
>>> I believe if we do this we will have fast initial load and the images
>>> (grayscale and rgb) will 'look right' when first loaded (i.e with good
>>> contrast) and it would not be necessary to assign any color value to
>>> grayscales by default.
>>>
>>> Regards
>>
>> Hi,
>> just my 2 ct.
>>
>> I guess the more important capability with raster is to open they faster
>> possible.
>>
>> I fear calculate all these values on first opened take more time than what
>> is waitable for the open of a raster.
>>
>> Perhaps this action should be choosable by settings ?
>>
>> Of course I see our situation.
>>
>> Our situation is of more sets of raster (7-800 each set) where every raster
>> is grey level or true color of 3-400 Mbyte each.
>> And all these raster are accessible by a remote shared server to our users.
>> For a total, actually, of about 13.000 rasters.
>>
>> I don't understand if this new feature could be really usable in a situation
>> like our.
>>
>> --
>> -----------------
>> Andrea Peri
>> . . . . . . . . .
>> qwerty àèìòù
>> -----------------
>>
>>
>> _______________________________________________
>> Qgis-developer mailing list
>> Qgis-developer at lists.osgeo.org
>> http://lists.osgeo.org/mailman/listinfo/qgis-developer
>>
>



More information about the Qgis-developer mailing list