It seems impossible since most of analyses are based on 3x3 cell neighborhood but I would like to hear your experiences. I know that some hydrodinamical models like TELEMAC are able to perform this.
You could perform an image re-sampling analysis on ArcGIS by using the high resolution image to enhance (to an extent) the resolution of the low/coarse image.
In addition, you could make use of the Climate Data Operators (CDO). The name of the function is cdoremapnn.
Thanks. But your suggestions are still related to uniform pixel size. I want to combine 2 or more different cell size within the same raster. Meanwhile, I found this article, which is very useful Article The Multiresolution Image Format
Perform registration of all rasters to the same reference system as the raster image. Then use a proper interpolation method (IDW, Spline, or Kriging).
A "long time ago" our team tested quadtrees for the representation of heterogeneous images, providing a spatially variable resolution which is related to the informativity of the image. See for example: https://ntrs.nasa.gov; Article Optimal tiling of heterogeneous images
) . You may find some useful ideas with 'quadtree' keyword as well as some new result, which may prove to be applicable to solve your questions.