Bioluminescence tomography (BLT) is an optical molecular imaging technique with high sensitivity and specificity, and it can provide three-dimensional distribution of the internal source according to the detected boundary light intensity. However, the source reconstruction with limited measurements is a challenging problem. We take advantage of the sparsity feature of bioluminescence source and formulate the BLT source reconstruction into an L1 norm minimization problem. The iterative detection support (ISD) algorithm is used to realize more accurate and faster reconstruction with limited data. The reconstruction algorithm alternatively runs two operations, i.e., support detection and signal reconstruction, until the solution meets the accuracy requirement. Simulations based on a digital mouse are designed to assess the location ability of the ISD method, and the result is compared with those of other three representative sparse algorithms. The simulation results show that the ISD method can achieve accurate reconstruction in both single-source and double-source cases.