The technique of detecting multiple dim and small targets with low signal-to-clutter ratios (SCR) is very important for infrared search and tracking systems. In this paper, we establish a detection method derived from maximal entropy random walk (MERW) to robustly detect multiple small targets. Initially, we introduce the primal MERW and analyze the feasibility of applying it to small target detection. However, the original weight matrix of the MERW is sensitive to interferences. Therefore, a specific weight matrix is designed for the MERW in principle of enhancing characteristics of small targets and suppressing strong clutters. Moreover, the primal MERW has a critical limitation of strong bias to the most salient small target. To achieve multiple small targets detection, we develop a hierarchical version of the MERW method. Based on the hierarchical MERW (HMERW), we propose a small target detection method as follows. First, filtering technique is used to smooth the infrared image. Second, an output map is obtained by importing the filtered image into the HMERW. Then, a coefficient map is constructed to fuse the stationary dirtribution map of the HMERW. Finally, an adaptive threshold is used to segment multiple small targets from the fusion map. Extensive experiments on practical data sets demonstrate that the proposed method is superior to the state-of-the-art methods in terms of target enhancement, background suppression and multiple small targets detection.