I am not super-frequent Photoshop user (CS5), so maybe this functionality already exists, and I just not aware of that. I am a frequent user of the GIS (mapping) software suite called ArcGIS though - and there I have a handy tool that is called 'Euclidean Allocation' that one can apply on images. It only supports one-channel images though. This page describes the functionality...
I am interested in hearing if there is an easy way to do this in Photoshop... One way to do something similar is using 'content-aware fill' but it is not exactly the same...
Here is an image where I have demonstrated the euclidean allocation functionality in ArcGIS:
The top is the original image, and the lower one is where I have applied this functionality. This way I can "grow" and "expand" the colors into the white (no data, transparent)
I don't think there is a built-in functionality in Photoshop that would do any such calculations out of the box. You may be able to do something with scripts, though. Maybe some of the developer pros here will have some ideas for you.
There's a ton of Voronoi filters out there based on pseudo-random distribution, but I'm not aware of an analytic Euclidian filter that could use real data either for PS or any other Adobe program...
In GIS it's usually used to determine pollution and what influence neighboring different types of area use have (usually in combination with other algoritms/ methods and run in multiple iterations) as well as determining safety perimeters as required for e.g. industrial buildings.... This may merely be a random example.
Mylenium - I am not interested in using real data, only the colors
Noel - I am a cartographer, and sometimes a thematic dataset (covering a land area) doesn't match my coastlines. To make the maps really pretty, I can use this functionality to "grow" the thematic dataset out into the ocean, and then I use the coastlines as a mask to not show the values in the ocean.
Please don't get to hung up on usage of data or GIS - I just want to "grow" the colors (without any mixing/aliasing) straight out into transparent pixels (or a selection) - a bit similar to 'content-aware fill' but in a way more simplistic...
Europe, Middle East and Africa