In this chapter, the authors propose an improved dynamic programming (DP) based track-before-detect (TBD) to detect small targets in a high pulse repetition frequency (PRF) radar. Unlike most DP TBD methods in the literature which only use the accumulated energy to decide if a target is present, the proposed algorithm jointly process several low-threshold-crossing data frames, and uses both the accumulated energy and the frequency of being associated to decide if a target is present.