Low-density parity-check code is a kind of near-optimal error correction code. Developing a highly efficient decoding algorithm is the main challenge of the implementation of LDPC code. In this paper, based on the min-sum column-layered decoding algorithm, a new selective computation technique is proposed to reduce the number of comparisons. With the presented decoding algorithm, the total number of floating number comparisons can be reduced dramatically (range from 81% to 91%) at the cost of negligible hardware overhead. Simulation results show that the decoding performance and iteration times are almost the same as the original column-layered decoding algorithm.