Abstract: In the field of sparse signal processing, compressive sensing is able to make signal sampling at a lower sampling frequency, while its reconstruction problem is equivalent to a special optimization problem that can be solved by the particle swarm optimization algorithm. For the purpose of compressive sensing, we firstly improve the particle swarm optimization algorithm by getting a theoretic lower-bound for the swarm size and proposing a new type of typology structure for the neighborhood and a multi-group collaboration learning paradigm and then construct the improved particle swarm optimization algorithm based compressive sensing reconstruction method especially applying for 2-D sparse signals. Finally, it is demonstrated by the experimental results on synthetic and real-world datasets that this proposed compressive sensing reconstruction method is feasible and effective.