PDF version of this document
This section summarises some of the previous preliminary developments; these are developments which were made before recent experiments to be listed in Chapter cha:Experiments of this report.
The simple data used by Smith was proving slightly too cumbersome for responsive experimentation on a relatively strong machine (1.8 GHz, 512MB RAM), especially owing to the complex algorithms devised for group-wise registration. It was at that point advisable that evaluation via profiling toolkits was made to hasten the process as much as possible. Alternatively, coding of the algorithm in a compiled language as C++ was seriously looked at as a possibility. The complexity of the departmental VXL library was believed to make a step as such less than desirable and no such development has ever been made thus far.
Once speed-up had been taken care of or when it was at least known that a nearly flawless well-performing piece of software was at the user's disposal (and one which was under control), the simple 1-D data could see the addition of a few additional characteristics. That new compositeB.6 data had to retain some good commonality and similarity across the set of images and it could not be overly more complex and unpredictable in comparison with a simple bump. A double-humped curve, a round smooth line or even a contour of a a profile of a face could be sensible and more challenging choicesB.7. In any case, whichever synthesis of data was eventually selected and experimented with, the choice of control points for the warping then became a more crucial issueB.8. A more localised control via warps then turned into a mandatory one because several separate structures exist in the data.
The experiments of Marsland, Twining and Taylor had already shown the realistic application of warps to a medium-resolution two-dimensional data. Nonetheless, it was vital to point out that an elliptical shape was dealt with and a priori knowledge of the problem was used to increase the speed of the group-wise registration process. Control points that characterise the warps were initially placed on a circle whose centre was the image centre and radius corresponded to the typical position of the skull in standardised imaging. If the problem involved point selection for, let us say, knee cartilage and no knowledge about the object was available in advance, the results would have then taken far more than 10 hours to obtain (as was the case for the 12 points distribution around the skull's exterior). Edge detection is quite useful in an application of this kind. It was highly useful in the case of the skull data, but finding edges that form a circle (confer Hough transform) as in a skull is somewhat of a simplified problem. Subsequent developments should aim to address many issues exhibiting resemblance to aforementioned ones.