The paper presents the BMB methodology for synthesis of binary functions, which can be applied to synthesis of functions implemented in LUT-based FPGAs. The BMB synthesis uses the functional decomposition based on symbolic decomposition and symbolic encoding. The methodology uses two optimization methods. First, the symbolic decomposition is used to build the multiple-valued network of functions. Second, the symbolic encoding is used to find an encoding that simplifies the decomposition. The symbolic optimization overpasses limits of the existing decomposition methods, where optimal encoding for LUT-based implementations is not possible. Simultaneously, the method can be used in the synthesis of high-level architecture where dependencies can be described using symbolic values. Experimental results showed that the presented methodology allows significant reduction of the cost of the FPGA implementation.