Normalization of relations on relational databases can avoid the redundancies of data and update anomalies, and it emphasizes the lossless decomposition of relations. The outcome of decomposition depends on the definition of the redundancies of tuples. In Extended Possibility-Based fuzzy data model, if the resemblance between the tuples exceeds the threshold, we can say that they are redundant tuples, but the relation of the resemblance doesn’t have the transitivity, so that there is no way to divide the tuples into non overlapping groups, and to merge the tuples in the same group. But, in the Similarity-Based fuzzy data model, the resemblance relation of redundant tuples has transitivity, and when merging the tuples, there is no previous problem. This study just discussed how to use of the definitions of calculating the resemblance between tuples and the definitions of fuzzy functional dependencies up to now, and then find out a way to lossless decompose the relations in Extended- Possibility and Similarity-Based fuzzy data model.