In this paper, we propose a two-step bit-transform-based normalized partial distortion search (TSB-NPDS) algorithm for fast motion estimation by using the characteristics of pattern similarity matching errors. Two significant bit-plane features of image block pattern extracted by a two-bit transform are used to determine the calculation order in the proposed TSB-NPDS. The experimental results indicate that the TSB-NPDS can achieve a speedup ranging from about 14.4 to 16.0 times of that of the full search algorithm with neglected PSNR loss. Furthermore, the performance of TSB-NPDS is better than other modified NPDS-based algorithms.