How to extract (x,y) coordinates of the boundary of a .tif image?

[Skaperen]

if i were to do this based on knowing nothing more, i'd take the RGB color image and make a monochrome image, if color is what i get.  i'd pick some value like 0xb0 and use that as the black (anything less) vs. white (anything greater) test. then i'd trace around looking for pixels with neighbors of the opposite value.  this makes for a 2-pixel boundary.  i don't (yet) know what to do (next) to get a (correct) 1-pixel boundary.  it probably depends on what the boundary is to be used for and how it will be used to make these decisions.

[merlem]

It's propably not fair to ask whether you throwed a glance to the code I've postet...
Only that I stayed in the monochromatic range until now. well, okay, I also did the colour check more mathematically.

[Skaperen]

to be honest, i did not look at the code.   i regularly skip that unless it is very small.  i prefer to take a more abstract view for issues in things like this.  i'd look at code if the issue is about something like a code expression not doing what is expected.

[merlem]

First, the RGB-type for the human pattern recognition part:

rgbimage = imac.convert("RGB")
areacol = rgbimage.getpixel((imsize[0] - 1, imsize[1] - 1))

for provenpix in provenpixels:
    rgbimage.putpixel(provenpix, (0, 200, 200))
for stillunprovpix in remainingunclear:
    rgbimage.putpixel(stillunprovpix, (200, 200, 0))
  
for stillunprovpix in remainingunclear:
    rgbimage.putpixel(stillunprovpix, (255, 0, 0))
    rgbimage.show()
    print "Is this a border pixel? It has the coordinates: ", stillunprovpix
    while True: 
        pixelinfo = raw_input("yes (y) or no (n)? ")
        if pixelinfo in ("y", "ye", "yes"):
            provenpixels.append(stillunprovpix)
            rgbimage.putpixel(stillunprovpix, (0, 125, 125))
            break
        elif pixelinfo in ("n", "no"):
            rgbimage.putpixel(stillunprovpix, areacol)            # remove mark
            break

Second...As I said, the code does mostly what you propose.
Only this part I don't understand:

 it probably depends on what the boundary is to be used for and how it will be used to make these decisions.

If "1-Pixel-border" is only what I assume it to be, there is no difference depending on the usage. It's just a question of "is this pixel one that makes the demarcation between the both colour regions?" 

And if you like to think about patterns: How would you solve a region like this in universal rules? I was not able to get a "full" solution.
Bit of ASCII art, "x" is black (background color) and "-" is blue (area colour):

xxxxxxxxxx-
xxxxxxx----
---xxxxx---
--xxxxxxx--
--xxxxxxx--
---xxxxxx--
-----------

With checking a maximum of 8 neighbours, I was able to prove the border state of a couple of pixels, shown here as "o":

xxxxxxxxxxo
xxxxxxxoo--
oooxxxxxo--
--xxxxxxx--
--xxxxxxxo-
---oxxxxx--
----oooo---

Switching to "o" for the pixels with unproven border state:

xxxxxxxxxx-
xxxxxxx--o-
---xxxxx---
--oxxxxxxo-
--oxxxxxx--
---xxxxxxo-
--------o--

I fear one has to widen up the range of neighbours to a total of at least 15 for those...