The estimation of backscatter coefficient, from in vivo tissue requires accurate compensation for intervening tissue attenuation. Current attenuation compensation methods (ACM) have been tested in fully or piecewise homogeneous phantoms. Thus, evidence of their performance in complex media remains scant. In this study we compare the performance of two ACM in a tissue-mimicking material with strong reflectors (SR). A gel-based phantom with SR was scanned with a L11-5v transducer on a Vantage 128 system (Verasonics). Estimates of from the phantom's background were obtained from the IQ echo signals using the known phantom attenuation (KA) or two ACM (a Constrained-Log-Difference (CLD) and a dynamic programming (DP) regularized method).The former method was used as gold standard. To analyze bias versus depth, a Pearson correlation coefficient was computed and the mean difference (MD) vs. depth was estimated. CLD-based showed a strong correlation (r = 0.96) with depth, while the correlation of DP-based estimates was weaker. DP-based values were closer to those obtained with KA (MD [0.9–1.15]) compared to the CLD-based estimates (MD [1.0–1.79]). The DP, regularized ACM show less sensitivity to the presence of SR compared to conventional ACM. We are currently using DP-based ACM to characterize breast lesions.