Superconducting quantum interference device (SQUID) can detect weak cardiac magnetic signals. By analyzing those detected cardiac magnetic signals can we provide the evidence for the diagnosis of cardiac diseases. In this paper, we reconstruct the cardiac current dipole array using minimum-norm least-squares method (MNLS), thus realize the imaging of equivalent cardiac current sources. In the process of doing inverse computation using MNLS for current dipole array reconstruction, we assume that the cardiac current sources are distributed in a plane layer in human heart and the cardiac magnetic signals needed for inverse computation are obtained with three methods: the simulation based on single current dipole and current multipole models, and the realistic measurements by SQUID. Using the three methods for obtaining the cardiac magnetic signals, we discuss the distributive characteristics of current dipoles from the inverse computation. Besides, we also add the constant noise and random noise to the cardiac magnetic signals and consider their influences on the reconstruction of current dipole array.