A binary code of length n is called a balanced code if each codeword contains exactly [n/2] (or [n/2]) ones and [n/2] (or [n/2]) zeros. In this paper, we give two improved methods for encoding and decoding balanced codes. The first one, called improved single map, improves the computation complexity of Knuth's single map function. This method, instead of complementing one bit at a time as done in Knuth's method, complements several appropriate bits at a time. Some simulation results show the improvement of this scheme over the previously known methods. The second one is a parallel implementation of this method.