This paper presents a novel algorithm for multiple subtraction based on independent component analysis (ICA) which takes into account the scale inconsistencies between the predicted and the true multiples. The proposed method avoids the unrealistic assumption in recent ICA based method that all the scale coefficients should be the same. Moreover, it settles the two ambiguities of the separated primary that are inherent in ICA by using some special characters of the mixing matrix. Synthetic seismic data set derived from the time-distance equation is applied to validate the effectiveness of the proposed algorithm, and experimental results demonstrate the improved performance of our algorithm