The principles of Independent Component Analysis (ICA) and factor analysis are given, and that the essence of ICA is factor rotation is presented. The analyses show that conventional factor rotations such as Varimax and Orthomax are equivalent conditionally to the early kurtosis-based estimate method for ICA presented. A Varimax-based method for ICA is proposed in consideration of all non-Gaussian sources even mixed signals with sub-and super-Gaussian distributions are included. Experimental results show that the novel method is simple and efficient when the mixing matrix is sparse.