I have a floar plan image (any shopping mall image) . I have to implement the direction for this image. like we shows directions on google map. In gaming we used Tiled Based system .
Now suppose user is at Food court he wants to go to Loading Dock need to show directions in same way that google map shows. Please suggest me a way to implement this.
If you can convert your map into a set of navigable polygons there are a number of good resources for navigation mesh generation (http://critterai.org/projects/nmgen_study/ and its references might be a good start). That's a somewhat tricky problem but a well defined and studied one.
I think you will find converting arbitrary images into machine readable maps much more difficult, if not impossible. You might use edge detection to split up the image into regions which can then be broken down into convex polygons but that's far from a complete solution. A line might represent a wall or it might be the edge of the "F" in "Food Court". Existing questions like https://mathematica.stackexchange.com/questions/19546/image-processing-floor-plan-detecting-rooms-borders-area-and-room-names-t take a different approach which might work better for some images. I think you need to set some constraints on what your images contain or at least build a larger sample set before you can determine how to even try to parse them successfully. Even then this becomes a hard image processing problem without a "right" answer. Instead expect to spend a lot of time refining a solution to generate less wrong results for a given set of training images.
I think this is, unfortunately, not a very good question for stack overflow because it will not get a specific answer.
