You may want to use this format when your algorithm process image computations using floating point arithmetic (continuous scheme). You may also use this format to represent any kind of two-dimensional real data (such as matrix).
Notice that you may lose precision when you use such format as the input of another module which requires integer representation (Cimage type), e.g. printing or displaying devices. In the other side, a module which accepts Fimage type as the input will also work without degradation if you put a Cimage type instead. It is so better to use, if possible, Cimage type for output variables and Fimage type for input.