The work of Agrawal et al. on the type of data mining known as association rule mining has been the basis for continuous research over the past seven years. Kuok, Fu and Wong have extended association rules with the fuzzy set theory of Zadeh to build a system more adapted to real world data. Meanwhile Bloch has recently applied this same fuzzy set theory to the area of image analysis, particularly to the task of finding relationships between image objects. Koperski and Han have been the leaders in adapting general data mining techniques, including association rules, to the domain of spatial data. This thesis describes a synthesis of these techniques to form a unified system for generalized image analysis, specifically finding general fuzzy rules about the relationships of objects in image data sets. Experimental results on synthetic data and real world data samples demonstrate that such a system is effective in producing interesting rules.