Accurate particle size characterization of aggregated and agglomerated particles is only possible by analysis of photographs. Both the primary particle size and the morphology of the aggregate are important experimental results. Since standard image analysis techniques for particle size analysis usually recognize only single particles, a new programme, called here Sparse Hough Transformation, was developed for the automated recognition of spherical particles within an aggregate. The method is shown to perform well, even for images with many overlapping particles. The structure of the aggregate is analysed using the fractal dimension, determined from the density-density correlation function. Finite size effects, important when dealing with aggregates containing few primary particles, are taken into account by including a cut-off function.