To seek a decoding scheme with good performance, low complexity and fast convergence, we present an improved parallel weighted bit flipping (IPWBF) algorithm for finite geometry low-density parity-check codes whose parity check matrix is of heavy row and column weights. In the IPWBF, a bit flipping (BF) function and two parallel BF criteria, all of which scatter in the literature, are exploited jointly to serve our purpose. Meanwhile, differential evolution is used to optimize the involved parameters. Simulation results show that the proposed algorithm achieves an observable performance gain over its counterparts without any complexity penalty. Furthermore, with respect to other known low complexity decodings such as normalized BPbased, the IPWBF yields a new performance versus complexity tradeoff, that is, higher throughput at the expense of moderate performance loss.