x-ray image processing, segmentation and wire mapping

Hey guys,

I am currently looking at how to speed up digitisation of a set of x-ray images (eg. http://imageshack.us/photo/my-images/341/lp5c10.png/). The goal is to semi-automate the process of using 9 points to describe the profile of each of the 16 wires seen in the picture. Please bear in mind that I am completely new to image processing as of a few days ago.

This is my best result so far in reducing the image. http://imageshack.us/photo/my-images/196/reduced.jpg/. Using a combination of average and Wiener filters, subtracting background, gray thresholding and skeletonisation. Unfortunately if you compare with the original image, you can see that some of the wire has been lost too. Advice regarding how best to reduce the image would be great.

Which leads to the second part.. I can see the lines but I don't really know how to get the computer to describe them. Using anaskel (a function somebody uploaded here) intersections and ends can be roughly identified (seen here http://imageshack.us/photo/my-images/526/wireybug.jpg/). The problem is, I still can't think how to bundle these points into separate groups that describe a line.

Really any methodology advice would be much appreciated!

Cheers, Simon

[Image Analyst]

I don't know what you're doing. The image is already digitized - how does all that stuff speed up digitization? I'm not sure which 9 points to get in that grid of wires and why you even want them or want to segment out the wire grid. Again, that has nothing to do with speeding up digitization of the image, which is already digital anyway.

[Walter Roberson]

"Digitization" is a term used especially in GIS, to indicate selecting points and finding the coordinates of them, and by extension, the path of entire lines.

[Simon]

whoops, i made the mistake of thinking i'd get an email for replies to this question - sorry.

alright, i'll try to clarify the problem. based upon the top image, current analysis methods require each of the 16 wires to be reconstructed. there are 8 vertical wires and 8 horizontal wires and the user clicks 9 points along each wire [from one end to the other], one by one. the 9 points representing each wire then have a spline fitted through them to approximate the full profile of the wire.

this is a time consuming process, as you must click 9*16 = 144 points total, and each point should be clicked as accurately as possible. so my goal is to speed it up somehow, like by clicking the start point and the end point of a wire and getting the full curve in between. i got closer since i posted this, so i put up a pic of it if that helps you understand what i'm trying to do [the concept worked but the result was kinda bad and the method is fragile].

Cheers,

Simon