[Gdal-dev] New Resampling Algorithm

[Julien Demaria]

Hi,

I want to implement an utility to reduce resolution of an image by 
averaging all input pixels contributing to one output pixel,
and I need use exactly the real contribution of each input pixel (may be a 
part of it if it is at the border of the input computed window 
corresponding to the output pixel coverage).

I've taken a look on the GDAL Warp API and the GDALWarpKernel class, to 
know if I can implement this as another resampling algorithm, as bilinear 
or bicubic,
but I'm not clear about the input pixels contribution : can I know in the 
resampling algorithm what are inputs pixels corresponding to one output 
pixel and how they contribute ?
I understand the density concept (float * 
GDALWarpKernel::pafUnifiedSrcDensity) is related to this, but not for one 
pixel, it is for all the chunk, isn't it ?

Do you think I can/should implement this algorithm as another resampling 
algorithm of the Warp API, or I've to develop a specific tool ? (note 
reprojection in't essential for me)


Thanks in advance for help,

Julien