This paper proposes three methods for reconstructing magnetic-source biocurrent distribution. These methods are more effective than the conventional pseudo-inversion-based reconstruction when the signal-to-noise ratio of measured data is low. First, a method of estimating magnetic- source-current covariance matrix using the measured-data covariance matrix is presented, and an averaged current squared-intensity distribution is reconstructed using the diagonal terms of the covariance matrix. The use of its off-diagonal terms leads to the second method that can separate magnetic-source activities correlated to each other from the uncorrelated activities. The third method is the Wiener reconstruction of current distributions based on the estimated source covariance matrix. Results of computer simulation demonstrate the effectiveness of those three methods.