What do you mean by fitting? It implies that you can somehow model or parametrize (i.e. calculate) the continuous shape and then adjust the parameters for best fit to the measured discrete dataset. You don't fit "the data", you fit some calculation to the data.
What you're saying is that you have 2 different measurements of the same thing, one discrete and one continuous (which is apparently less precise), but what do you want to achieve? An example would help... If you want to somehow alter the continuous measurement so it fits to the discrete data points, you still need some kind of parametrized model, i.e. calculate the 2D shape as a function of some parameters, or know how the continuous measurement relates to the discrete one. If it's a simple geometric relationship (e.g. linear/nonlinear scaling, rotation, warping etc.), you can use the corresponding parameters to manipulate the continuous data for best fit with the discrete data and noise can be reduced by low-pass filtering (or applying a smoothing kernel in case of equidistantly sampled 2D data). However, if you don't know how to parametrize the relationship between the two measurements (or the nature of the error of the continuous measurement), you're out of luck without a proper theoretical model and all you can do is interpolation between the discrete measured data points (in your case probably "interpolate 2D scattered").
Nice username btw...
Search LabVIEW like a graph! 
