| This research emphasizes the development of a novel technique for 3-dimensional imaging of fat crystal networks. A model fat system was crystallized by conventional methods (a 2D thin film of 20mum in thickness) and in a volume of approximately 200mum in thickness. 3-Dimensional data sets of optical slices through the volume were then deconvolved by a nearest neighbours (NN) algorithm and a blind algorithm. NN deconvolution was not successful in removing out-of-focus pixels, however blind deconvolution was highly successful, as was demonstrated by 3D renderings and 2D projections of the volumes. Blind deconvolved images of slices and 2D projections of the volume were analysed and compared to results for conventional thin film crystallization. Image analysis of the volume slices and projections had a lower area fraction, lower particle counting fractal dimension (Df), and smaller mean particles sizes, indicating that confinement effects in thin films results in more orderly crystallization. The fractal dimension Db, on the other hand, was not sensitive to confinement. Fractal dimension retention upon projection of 3D slices onto a 2D plane provided novel experimental evidence for this theorized behaviour. Constancy in fractal dimension values was demonstrated under variable imaging criteria including microscopy optics, objective magnification, type of camera, and image size and resolution. A novel means of determining the relative degree of order in particle distributions by analysis of nearest neighbour distances was also successfully demonstrated. |
