Re: [Openev-discuss] OpenEV Rasters Types

[Andrey K. <dr...@ak...>]

On Tue, Jun 29, 2004 at 09:58:05AM +0200, Julien Demaria wrote:
> I use int32 rasters in OpenEV and I've observed that the pixels values
> I see in the tracktool are truncated. It seems that the reason is this
> code in gvraster.c which read a pixel in my int32 layer as a float,
> and the cast losts informations :
> 
> "...
> float   value[2];
> 
> GDALRasterIO( raster->gdal_band, GF_Read, pixel, line, 1, 1,
>                         value, 1, 1, GDT_CFloat32, 0, 0 );
> 
> *real = value[0];
> *imaginary = value[1];
> ..."
> 
> 
> I've looked in gvraster.c and I understand that rasters are classified in 
> two types : Byte and others, and others have all the same GDAL GDT_Float32 
> type...
> Are there reasons for this ?
> Does it mean that (u)int32 and float64 GDAL rasters aren't well
> handled by OpenEV ?
> 
> gvrastertypes.h defines types for short and int rasters, but they aren't 
> used, and types for ushort, uint and double rasters aren't defined.

Julien,

I don't see any reason why we can't get GDT_CFloat64 values from GDAL,
especially if we recall that *real and *imaginary are double precition
floats and we have unnecessary type conversion here.

Could you change definition for value[] arry to double and replace
GDT_CFloat32 with the GDT_CFloat64 and test whether it helps for your
data?

							Andrey

-- 
Andrey V. Kiselev
Home phone:  +7 812 5274898  ICQ# 26871517