Step 1) Administrative / data entry worker opens ?, loads raster image, and using a 'lasso-like' tool (a la Photoshop 'magnetic' if possible) specifies areas of interest / masks.
Step 2) ?
Step 3) Profit! (i.e., application opens raster image, vector file describing masks, does interesting stuff to areas of interest based on runtime context)
My question is 'what do we use for steps 1 and 2?'
Primary goal is to reduce the time required to complete Steps 1 and 2 (the 'mechanical Turk' part of the process). Strictly secondary goal is the ease of parsing / using the vector format describing the areas of interest / masks in step 3.
Water cooler conversation: 'Sheesh, there's got to be some Flash component that does this!" "What about GIMP?" "If we have to put C or C++ programmers on it, it's not going to be a good investment...." "But what if it only takes a day or two?" "Nothing takes just a day or two," ....
For prototyping, the obvious route is Photoshop -> Illustrator -> SVG. But if there's a framework / library / editor that could make the 'lasso' coding and vector-file generation easy, it might tip the scales in favor of writing a custom editor.
Is there such a framework / library / editor?