Abstract: A new method of selecting reference vector for fast search algorithm using inequality in subspace proposed by Cui et al is proposed.Before the searching process,all codewords are transformed in Hadamard-transformed and sorted in an ascending order according to their first dimensions.The sorted codewords in the codebook are divided into four groups equally and the middle codeword of each group is selected as the reference vector.The method of selecting reference vector in this paper can increase the ability of fast search algorithm using inequality in subspace in rejecting the unlikely codewords and reduce the searching space and speed up the searching time.The experimental results show that the proposed method has a better performance than other fast search algorithm in searching time for different test images.