Since there is no good answer to this I'll add one now.

The easiest way of finding the centre of a Siemen's star is to detect the edges of the vanes (all or most of them) and trace these back to the centre. If you have a tight crop of the star, you could do this by edge detecting and then doing a Radon transform on the edge detected image. This should produce high values were the lines lines are inline with the axis of the sum.

You should then have several angles in the Radon transform with high results and can the work out the intersection of the lines easily.

You could also skip out the Radon transform and directly fit to the lines and find the intersection of several straight lines.

These methods all assume that the lines are actual lines that are straight, some kinds of distortion might cause issues with this.

The above methods should also be robust to rotation and scale and so some additional image content. They should also be robust to defocus and other image effects that you might want to observe in the star - since they use the large scale information as well as the small scale information.